FINAL REJECITON
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This action is in response to the amended application filed on September 16th 2025 and the original application filed on June 16th, 2022. Acknowledgment is made of applicant's claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. IL277910, filed on October 9th, 2020.
Response to Amendment
The Examiner thanks the applicant for the remarks, edits and arguments.
Regarding Claim Rejections – 35 U.S.C. § 112
Applicant Remarks:
The applicant has made amendments to claims 1, 20 and 21 they believe the claims no longer recite indefinite language. Therefore, the rejection under 35 U.S.C. § 112 should be withdrawn.
Examiner Response:
Upon evaluation of the amended claims the examiner has still found issues pertaining to claims 1, 20, and 21. In particular the limitation from claim 1, “receiving, from each second computing system, an encrypted second secret share of the inference result; and”. It is still unclear where and how the second inference result is produced. It is unclear whether the second computing device is generating the second inference result concurrently and if that second device will rely or use information from a third or fourth device. As the claim reads, it appears that a first system on relies on other systems to produce a result, where the other systems rely on more systems to produce their own result. This is a cascade affect and one system would rely on n+1 systems to produce a single result. Therefore, because it is unclear how this second inference result is generated and claims 1, 20 and 21 are still rejected under 35 U.S.C. 112(b) and the rejection is upheld, see 112 rejections below.
Regarding Claim Rejections – 35 U.S.C. § 101
Applicant Remarks:
The applicant has made amendments to claims 1, 20 and 21 they believe the claims recite patent eligible subject matter. Therefore, the rejection under 35 U.S.C. § 101 should be withdrawn.
Examiner Response:
After each amendment the examiner will evaluate the amended claims to see if they disclose patent eligible subject matter under 35 U.S.C. § 101. Upon reconsideration of the amended claims, it was found that they still recite abstract ideas and therefore recite patent ineligible subject matter. Regarding claim 1, it was found that the claims recites the abstract ideas of: predicting a user based on data, determining a residue value, and generating a inference result. It is understood that a human, with the assistance of pen and paper or a computer as a tool, could evaluate data and produce an inferred result based on that evaluation. After determining the use of an abstract idea in the claims the examiner must continuing the Alice/Mayo test and see if the claims as a whole recites more than a judicial exception and if there are significant improvements to designated fields. While performing this step the examiner has evaluated the claims as a whole and it was still found that the claims recite subject matter relating to well-understood or routine concepts such as transmitting data over a network and/or providing instructions. Therefore, the rejection under 35 U.S.C. 101 has been upheld, see 101 rejection below.
Regarding Claim Rejections – 35 U.S.C. § 102 / 103
Applicant Remarks:
The applicant argues that claim 1 has been amended and the art Shukla fails to teach or disclose the claim. Further the applicant states that Shukla in particular fails to teach the limitations in claim 1 and that Shukla fails to disclose the use of a machine learning model. Because of this Shukla and any combination of arts with Shukla would fundamentally not teach he claimed limitations. Finally, the applicant states that claims 19 and 20 recite similar subject matter to claim 1 and it is argued that these claims should be allowed for the same reasons as claim 1. Because of these reasons the applicant has requested the rejection under 35 U.S.C. § 102/103 be withdrawn.
Examiner Response:
After each submitted amendment, the examiner will evaluate the amended claims and see if the art still teaches the disclosed subject matter. Upon further examination it was found that Shukla does fail to teach the amended claims. Further search was performed and the examiner did find new art which is able to better teach the amended claims. However, while searching there was no single art that was able to teach the independent claims. Therefore, the rejection under 35 U.S.C. § 102 has been withdrawn. With that, it was found that a combination of art does appear to teach the amended claims. In combination of the previously provided art and Perez et al. the examiner believes it would have been obvious to someone of the art, before the filing date of this application, to reproduce the claimed invention. Therefore, the rejection under 35 U.S.C. § 103 is upheld, see 103 rejection below.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1, 20 and 21 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites the limitation, “receiving, from each second computing system, an encrypted second secret share of the inference result; and” (Emphasis added). It is not clear where the second inference result is produced by the second system. It is not clear whether the second computing devices is generating the second inference result concurrently and if that second device will rely on or use information from a third device or fourth device. As the claim reads, it appears that the system as a whole relies on other systems to produce a portion of an end result, meaning that one system will be require other systems to produce a result. This would cause a cascade affect and a final result would be difficult to achieve as it would require n+1 systems to complete. Therefore, it is not clear how the second inference result is generated, and this claim is rejected under 35 U.S.C. 112(b) for being indefinite. For examination purposes the claim will be interpreted to mean, “receiving, from each second computing system, an encrypted second secret share of the inference result … wherein the second result is generated by only one device and is for the first computing device to generate a final result …” Dependent claims 2-19 depend on rejected claim 1 and are also rejected under 35 U.S.C. 112(b) by virtue of this dependency. Appropriate correction is required.
Claim 20 recites the limitation, “receiving, from each second computing system, an encrypted second secret share of the inference result; and” (Emphasis added). It is not clear where the second inference result is produced by the second system. It is not clear whether the second computing devices is generating the second inference result concurrently and if that second device will rely on or use information from a third device or fourth device. As the claim reads, it appears that the system as a whole relies on other systems to produce a portion of an end result, meaning that one system will be require other systems to produce a result. This would cause a cascade affect and a final result would be difficult to achieve as it would require n+1 systems to complete. Therefore, it is not clear how the second inference result is generated, and this claim is rejected under 35 U.S.C. 112(b) for being indefinite. For examination purposes the claim will be interpreted to mean, “receiving, from each second computing system, an encrypted second secret share of the inference result … wherein the second result is generated by only one device and is for the first computing device to generate a final result …”. Appropriate correction is required.
Claim 21 recites the limitation, “receiving, from each second computing system, an encrypted second secret share of the inference result; and” (Emphasis added). It is not clear where the second inference result is produced by the second system. It is not clear whether the second computing devices is generating the second inference result concurrently and if that second device will rely on or use information from a third device or fourth device. As the claim reads, it appears that the system as a whole relies on other systems to produce a portion of an end result, meaning that one system will be require other systems to produce a result. This would cause a cascade affect and a final result would be difficult to achieve as it would require n+1 systems to complete. Therefore, it is not clear how the second inference result is generated, and this claim is rejected under 35 U.S.C. 112(b) for being indefinite. For examination purposes the claim will be interpreted to mean, “receiving, from each second computing system, an encrypted second secret share of the inference result … wherein the second result is generated by only one device and is for the first computing device to generate a final result …” Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-21 are rejected under 35 U.S.C 101 because the claimed invention is
directed to an abstract idea without significantly more. The analysis of the claims will
follow the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50
(“2019 PEG”).
Claim 1
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Claim 1, recites “A computer-implemented method comprising:” therefore it is directed to the statutory category of a process.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“executing, by the first computing system in collaboration with one or more second computing systems of the plurality of MPC computing systems, a first machine learning model using the given user profile as input data, wherein the first machine learning model is trained using a plurality of user profiles;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use a computer as a tool to perform actions such as using a machine learning model. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates a predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use a computer as a tool to evaluate data and determine a value based on accuracy metrics. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from one or more sources and compare the data to determine a value. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“generating, by the first computing system, a first secret share of an inference result based at least in part on the predicted label determined for the given user profile and the predicted residue value;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from difference devices and produce a result from the pooled data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by a first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“receiving, as an output of the first machine learning model, a predicted label for the given user profile;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“receiving, by the first computing system and from each second computing system of the one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“receiving, from each second computing system, an encrypted second secret share of the inference result; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by a first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“receiving, as an output of the first machine learning model, a predicted label for the given user profile;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“receiving, by the first computing system and from each second computing system of the one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“receiving, from each second computing system, an encrypted second secret share of the inference result; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 2
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“determining, by the first computing system, a first share of the predicted label based at least in part on (i) the first share of the given user profile, (ii) the first machine learning model trained using the plurality of user profiles, and (iii) one or more of the plurality of true labels for the plurality of user profiles, the plurality of true labels including one or more true labels for each user profile in the plurality of user profiles;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data and portion that data based on data factors. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining the predicted label based at least in part on the first and second shares of the predicted label.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from multiple sources and make a determination on that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by the first computing system and from the second computing system, data indicating a second share of the predicted label determined by the second computing system based at least in part on a second share of the given user profile and a first set of one or more machine learning models; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by the first computing system and from the second computing system, data indicating a second share of the predicted label determined by the second computing system based at least in part on a second share of the given user profile and a first set of one or more machine learning models; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 3
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“determining, by the first computing system, a first share of the predicted label based at least in part on the first transformed share of the given user profile.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from multiple source and determine a result with that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “applying, by the first computing system, a transformation to the first share of the given user profile to obtain a first transformed share of the given user profile, wherein determining, by the first computing system, the first share of the predicted label comprises:” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “applying, by the first computing system, a transformation to the first share of the given user profile to obtain a first transformed share of the given user profile, wherein determining, by the first computing system, the first share of the predicted label comprises:” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 4
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein the transformation comprises a Johnson-Lindenstrauss (J-L) transformation.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein the transformation comprises a Johnson-Lindenstrauss (J-L) transformation.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 5
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein determining, by the first computing system, the first share of the predicted label comprises: providing, by the first computing system, the first transformed share of the given user profile as input to the first machine learning model to obtain a first share of the predicted label for the given user profile as output.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein determining, by the first computing system, the first share of the predicted label comprises: providing, by the first computing system, the first transformed share of the given user profile as input to the first machine learning model to obtain a first share of the predicted label for the given user profile as output.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 6
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“evaluating a performance of the first machine learning model, comprising, for each of the plurality of user profiles: determining a predicted label for the user profile, comprising: determining, by the first computing system, a first share of a predicted label for the user profile based at least in part on (i) a first share of the user profile, (ii) the first machine learning model, and (iii) one or more of the plurality of true labels for the plurality of user profiles;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human ais able to evaluate user data from multiple sources to produce a result. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining the predicted label for the user profile based at least in part on the first and second shares of the predicted label;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human ais able to evaluate user data from multiple sources to produce a result. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining a residue value for the user profile indicating an error in the predicted label, comprising: determining, by the first computing system, a first share of the residue value for the user profile based at least in part on the predicted label determined for the user profile and a first share of a true label for the user profile included in the plurality of true labels;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human ais able to evaluate user data from multiple sources to produce a result. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining the residue value for the user profile based at least in part on the first and second shares of the residue value; and” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human ais able to evaluate user data from multiple sources to produce a result. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by the first computing system and from the second computing system, data indicating a second share of the predicted label for the user profile determined by the second computing system based at least in part on a second share of the user profile and the first set of one or more machine learning models maintained by the second computing system; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“receiving, by the first computing system and from the second computing system, data indicating a second share of the residue value for the user profile determined by the second computing system based at least in part on the predicted label determined for the user profile and a second share of the true label for the user profile; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“training the second machine learning model using data indicating the residue values determined for the plurality of user profiles in evaluating the performance of the first machine learning model.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by the first computing system and from the second computing system, data indicating a second share of the predicted label for the user profile determined by the second computing system based at least in part on a second share of the user profile and the first set of one or more machine learning models maintained by the second computing system; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“receiving, by the first computing system and from the second computing system, data indicating a second share of the residue value for the user profile determined by the second computing system based at least in part on the predicted label determined for the user profile and a second share of the true label for the user profile; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“training the second machine learning model using data indicating the residue values determined for the plurality of user profiles in evaluating the performance of the first machine learning model.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 7
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“before evaluating the performance of the first machine learning model: deriving a set of parameters of a function, comprising: deriving, by the first computing system, a first share of the set of parameters of the function based at least in part on a first share of each of the plurality of true labels;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data and determine parameters based on that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“deriving the set of parameters of the function based at least in part on the first and second shares of the set of parameters of the function; and” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data and determine parameters based on that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by the first computing system and from the second computing system, data indicating a second share of the set of parameters of the function derived by the second computing system based at least in part on a second share of each of the plurality of true labels; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“configuring the first machine learning model to, given a user profile as input, generate an initial predicted label for the user profile and apply the function, as defined based on the derived set of parameters, to the initial predicted label for the user profile to generate, as output, a first share of a predicted label for the user profile.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by the first computing system and from the second computing system, data indicating a second share of the set of parameters of the function derived by the second computing system based at least in part on a second share of each of the plurality of true labels; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“configuring the first machine learning model to, given a user profile as input, generate an initial predicted label for the user profile and apply the function, as defined based on the derived set of parameters, to the initial predicted label for the user profile to generate, as output, a first share of a predicted label for the user profile.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 8
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“estimating, by the first computing system, a first share of a set of distribution parameters based at least in part on the first share of each of the plurality of true labels, wherein deriving, by the first computing system, the first share of the set of parameters of the function based at least in part on the first share of each of the plurality of true labels comprises:” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate a set of data and produce an estimated result based on that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“deriving, by the first computing system, a first share of the set of parameters of the function based at least in part on the first share of the set of distribution parameters.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data and determine a set of parameters based on that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim 9
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein the set of distribution parameters include one or more parameters of a probability distribution of prediction errors for true labels of a first value in the plurality of true labels, and one or more parameters of a probability distribution of prediction errors for true labels of a second value in the plurality of true labels, the second value being different from the first value.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein the set of distribution parameters include one or more parameters of a probability distribution of prediction errors for true labels of a first value in the plurality of true labels, and one or more parameters of a probability distribution of prediction errors for true labels of a second value in the plurality of true labels, the second value being different from the first value.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 10
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“the first share of the residue value for the user profile is indicative of a difference in value between the predicted label determined for the user profile and the first share of the true label for the user profile; and” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to determine a value based on evaluating data from multiple sources. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“the second share of the residue value for the user profile is indicative of a difference in value between the predicted label determined for the user profile and the second share of the true label for the user profile.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to determine a value based on evaluating data from multiple sources. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
This claim does not recite any additional limitations which integrate the abstract idea into a practical application.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea and thus the claim is subject-matter ineligible.
Claim 11
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “the first machine learning model includes a k-nearest neighbor model maintained by the first computing system;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“the first set of one or more machine learning models includes a k-nearest neighbor model maintained by the second computing system;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“the second machine learning model includes at least one of a deep neural network (DNN) maintained by the first computing system and a gradient-boosting decision tree (GBDT) maintained by the first computing system; and” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“the second set of one or more machine learning models includes at least one of a DNN maintained by the second computing system and a GBDT maintained by the second computing system.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “the first machine learning model includes a k-nearest neighbor model maintained by the first computing system;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“the first set of one or more machine learning models includes a k-nearest neighbor model maintained by the second computing system;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“the second machine learning model includes at least one of a deep neural network (DNN) maintained by the first computing system and a gradient-boosting decision tree (GBDT) maintained by the first computing system; and” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“the second set of one or more machine learning models includes at least one of a DNN maintained by the second computing system and a GBDT maintained by the second computing system.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 12
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“wherein determining, by the first computing system, the first share of the predicted label comprises: identifying, by the first computing system, a first set of nearest neighbor user profiles based at least in part on the first share of the given user profile and the k-nearest neighbor model maintained by the first computing system;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data and determine similarities in the data and produce a result using that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“identifying a number k of nearest neighbor user profiles that are considered most similar to the given user profile among the plurality of user profiles based at least in part on the first and second sets of nearest neighbor profiles; and” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data and determine similarities in the data and produce a result using that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining, by the first computing system, the first share of the predicted label based at least in part on a true label for each of the k nearest neighbor user profiles.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data and determine similarities in the data and produce a result using that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by the first computing system and from the second computing system, data indicating a second set of nearest neighbor profiles identified by the second computing system based at least in part on the second share of the given user profile and the k-nearest neighbor model maintained by the second computing system;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by the first computing system and from the second computing system, data indicating a second set of nearest neighbor profiles identified by the second computing system based at least in part on the second share of the given user profile and the k-nearest neighbor model maintained by the second computing system;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 13
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“wherein determining, by the first computing system, the first share of the predicted label further comprises: determining, by the first computing system, a first share of a sum of the true labels for the k nearest neighbor user profiles;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from multiple sources and produce a result based on this data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining the sum of the true labels for the k nearest neighbor user profiles based at least in part on the first and second shares of the sum of the true labels for the k nearest neighbor user profiles.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from multiple sources and produce a result based on this data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by the first computing system and from the second computing system, a second share of the sum of the true labels for the k nearest neighbor user profiles; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by the first computing system and from the second computing system, a second share of the sum of the true labels for the k nearest neighbor user profiles; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 14
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein determining, by the first computing system, the first share of the predicted label further comprises: applying a function to the sum of the true labels for the k nearest neighbor user profiles to generate the first share of the predicted label for the given user profile.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein determining, by the first computing system, the first share of the predicted label further comprises: applying a function to the sum of the true labels for the k nearest neighbor user profiles to generate the first share of the predicted label for the given user profile.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 15
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein the first share of the predicted label for the given user profile comprises the sum of the true labels for the k nearest neighbor user profiles.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein the first share of the predicted label for the given user profile comprises the sum of the true labels for the k nearest neighbor user profiles.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 16
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“wherein determining, by the first computing system, the first share of the predicted label based at least in part on the true label for each of the k nearest neighbor user profiles comprises: determining, by the first computing system, a first share of a set of predicted labels based at least in part on a set of true labels for each of the k nearest neighbor user profiles corresponding to a set of categories, respectively, comprising, for each category in the set:” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from multiple sources and produce a result based on this data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining a first share of a frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of a first value;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from multiple sources and produce a result based on this data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value based at least in part on the first and second shares of the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value.” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from multiple sources and produce a result based on this data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by the first computing system and from the second computing system, a second share of the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by the first computing system and from the second computing system, a second share of the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 17
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“wherein determining, by the first computing system, the first share of the set of predicted labels comprises, for each category in the set:” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data and produce a result based on that data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements,
“applying a function corresponding to the category to the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value to generate a first share of a predicted label corresponding to the category for the given user profile.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “applying a function corresponding to the category to the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value to generate a first share of a predicted label corresponding to the category for the given user profile.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 18
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites the abstract ideas of the preceding claims from which it depends.
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein the client device computes the given user profile using a plurality of feature vectors that each include feature values related to events of a user of the client device and a decay rate for each feature vector.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein the client device computes the given user profile using a plurality of feature vectors that each include feature values related to events of a user of the client device and a decay rate for each feature vector.” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 19
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
A process, as above.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“classifying one or more of the plurality of feature vectors as sparse feature vectors; and” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate and classify data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “wherein the client device computes the given user profile using a plurality of feature vectors that each include feature values related to events of a user of the client device, wherein computing the given user profile comprises:” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “wherein the client device computes the given user profile using a plurality of feature vectors that each include feature values related to events of a user of the client device, wherein computing the given user profile comprises:” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 20
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Claim 20, recites “A system comprising: one or more processors of a first computing system; and one or more storage devices storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:” therefore it is directed to the statutory category of a machine.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“executing, by the first computing system in collaboration with one or more second computing systems of the plurality of MPC computing systems, a first machine learning model using the given user profile as input data, wherein the first machine learning model is trained using a plurality of user profiles;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use a computer as a tool to perform actions such as using a machine learning model. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates a predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use a computer as a tool to evaluate data and determine a value based on accuracy metrics. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from one or more sources and compare the data to determine a value. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“generating, by the first computing system, a first secret share of an inference result based at least in part on the predicted label determined for the given user profile and the predicted residue value;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from difference devices and produce a result from the pooled data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by a first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“receiving, by the first computing system and from each second computing system of the one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“receiving, from each second computing system, an encrypted second secret share of the inference result;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by a first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“receiving, by the first computing system and from each second computing system of the one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“receiving, from each second computing system, an encrypted second secret share of the inference result;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 21
Step 1 – Is the claim to a process, machine, manufacture or composition of matter?
Claim 21, recites “A non-transitory computer readable storage medium carrying instructions that, when executed by one or more processors of a first computing system, cause the one or more processors to perform operations comprising:” therefore it is directed to the statutory category of a machine.
Step 2A Prong 1 – Does the claim recite an abstract idea, law of nature, or natural
phenomenon?
The claim recites, inter alia:
“executing, by the first computing system in collaboration with one or more second computing systems of the plurality of MPC computing systems, a first machine learning model using the given user profile as input data, wherein the first machine learning model is trained using a plurality of user profiles;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use a computer as a tool to perform actions such as using a machine learning model. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to use a computer as a tool to evaluate data and determine a value based on accuracy metrics. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from one or more sources and compare the data to determine a value. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
“generating, by the first computing system, a first secret share of an inference result based at least in part on the predicted label determined for the given user profile and the predicted residue value;” Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of evaluating and observing data, which is an evaluation or observation that is practically capable of being performed in the human mind with the assistance of pen and paper. A human is able to evaluate data from difference devices and produce a result from the pooled data. The limitation is merely applying an abstract idea on generic computer system. See MPEP 2106.04(a)(2)(III)(c).
Step 2A Prong 2 – Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements, “receiving, by the first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“receiving, as an output of the first machine learning model, a predicted label for the given user profile;” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising at the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“receiving, by the first computing system and from each second computing system of the plurality of computing systems one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“receiving, from each second computing system, an encrypted second secret share of the inference result; and” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” is an insignificant extra-solution activity required for any uses of the mental processes (see MPEP § 2106.05(g)) As such, the claim is ineligible.
Step 2B – Does the claim recite additional elements that amount to significantly more than the judicial exception?
Finally, the claim taken as a whole does not contain an inventive concept which provides significantly more than the abstract idea. The additional elements, “receiving, by the first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“receiving, as an output of the first machine learning model, a predicted label for the given user profile;” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising at the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” amounts to generic computer components used as a tool to perform an existing process. Thus, the additional element amounts to no more than a recitation of the words "apply it" (or an equivalent) or are more than mere instructions to implement an abstract idea or other exception on a computer (see MPEP § 2106.05(f)).
“receiving, by the first computing system and from each second computing system of the plurality of computing systems one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“receiving, from each second computing system, an encrypted second secret share of the inference result; and” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” is an insignificant extra-solution activity required for any uses of abstract ideas (see MPEP § 2106.05(g)), and is a well-understood, routine, conventional activity (see MPEP § 2106.05(d)(i); “Receiving or transmitting data over a network, e.g., using the Internet to gather data”.
Taken alone or in combination, the additional elements of the claim do not provide an inventive concept and thus the claim is subject-matter ineligible.
Claim 22 (Cancelled)
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 2, 20, and 21 are rejected under 35 U.S.C. 103 as being unpatentable over Perez et al., (Perez et al., “You Are Your Metadata: Identification and Obfuscation of Social Media Users using Metadata Information”, May 2018, hereinafter “Perez”) in view of Montiel et al., (Montiel et al., “Privacy-preserving multi-party cohort discovery”, 2016, hereinafter “Montiel”).
Regarding claim 1, Perez discloses, “A computer-implemented method comprising:” (Introduction, pp. 1-2; “In this paper, we present an in-depth analysis of the identification risk posed by metadata to a user account. We treat identification as a classification problem and use supervised learning algorithms to build behavioral signatures for each of the users. Our analysis is based on metadata associated to micro-blogging services like Twitter: each tweet contains the metadata of the post as well as that of the account from which it was posted. However, it is worth noting that the methods presented in this work are generic and can be applied to a variety of social media platforms with similar characteristics in terms of metadata.” This article proposes a method which uses meta data to identify a user in an online social network. This is a method which uses machine learning to produce an output.)
“receiving, by a first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” (Attack Model, pp. 2; “In our evaluation, as discussed before, the input of the classifier is a set of new (unseen) tweets. We refer to a successful prediction of the account identity as a hit and an unsuccessful one as a miss. We assume that the attacker is able to access the metadata of tweets from a group of users together with their identities (i.e., the training set) and that the new tweets belong to one of the users in the training set.” This model will take in user data, which does not directly identify the user, and is able to determine the user. This is used with twitter which teaches the use of a large social network containing multiple computing systems.)
“executing, by the first computing system in collaboration with one or more second computing systems of the plurality of MPC computing systems, a first machine learning model using the given user profile as input data, wherein the first machine learning model is trained using a plurality of user profiles;” (Implementation of the Classifiers, pp. 3; “The optimization of the internal parameters for each classifier was conducted as a combination of best practices in the field and experimental results in which we used the cross validated grid search capability offered by SciKit-learn. In summary, we calculated the value of the parameters for each classifier as follows. For KNN, we consider the single closest value based on the Euclidean distance between the observations; for RF, we chose entropy as the function to measure the effectiveness of the split of the data in a node; finally, for MLR, we selected the limited-memory implementation of the Broyden-Fletcher-Goldfarb-Shanno (LM-BFGS) optimizer as the value to optimize (Liu and Nocedal 1989).” The model in this article uses machine learning to predict a user based on user metadata. This metadata is public, which teaches the use of machine learning to identify a user based on outstanding, non-personal identifying data.)
“receiving, as an output of the first machine learning model, a predicted label for the given user profile;” (Attack Model, pp. 2; “We present the likelihood of success of an identification attack where the adversary’s ultimate goal is to identify a user from a set given this knowledge about the set of accounts. To achieve this, we answer this question: Is it possible to identify an individual from a set of metadata fields from a randomly selected set of Twitter user accounts?” This article teaches a method which is able to identify a user based on public meta data. This will return a result stating the inferred user for the given data. This teaches the output of a result from a machine learning model.)
“generating, by the first computing system, a first secret share of an inference result based at least in part on the predicted label determined for the given user profile and the predicted residue value;” (Implementation of the Classifiers, pp. 3; “The optimization of the internal parameters for each classifier was conducted as a combination of best practices in the field and experimental results in which we used the cross validated grid search capability offered by SciKit-learn. In summary, we calculated the value of the parameters for each classifier as follows. For KNN, we consider the single closest value based on the Euclidean distance between the observations; for RF, we chose entropy as the function to measure the effectiveness of the split of the data in a node; finally, for MLR, we selected the limited-memory implementation of the Broyden-Fletcher-Goldfarb-Shanno (LM-BFGS) optimizer as the value to optimize (Liu and Nocedal 1989).” The model in this article uses machine learning to predict a user based on user metadata. This metadata is public, which teaches the use of machine learning to identify a user based on outstanding, non-personal identifying data.)
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” (Dynamic Attributes, pp. 6; “Table 5 presents similar results for the RF algorithm. Even without the ACT, we are able to achieve 94.41% accuracy in a group of 10,000 users. These results are directly linked to the behavior of an account and are obtained from a multiclass model.” Tables 4-7 disclose the results of the experiments conducted by this model. This represents the returned values of how accurate the model was at identifying a user based on different levels of user attributes. This teaches the final results of the system and how accurate it was.)
Perez fails to explicitly disclose, “receiving, from each second computing system, an encrypted second secret share of the inference result; and”, “executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates a predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:”, “executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;”, “receiving, by the first computing system and from each second computing system of the one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and”, and “determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;”.
However, Montiel discloses, “receiving, from each second computing system, an encrypted second secret share of the inference result; and” (Data Obfuscation Model, pp. 10; “As describe above, in case of an untrusted mediator, the parties further encrypt or obfuscate their data. We study two data obfuscation schemes in this work, Johnson-Lindenstrauss embedding and secure binary embedding. In both cases, the party’s will need to agree apriori on a common encryption key. The encryption key is secret to the parties, which will be used to decrypt the results back.” The model in this article discloses different models in a multiple party computation environment. Two models disclosed in the article described the use of untrusted mediators. A system using this proposed architecture uses encryption methods to transmit data from user to user or user to the mediator. Figure 6 shows this process of encryption and data transferring.)
“executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates a predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:” (Average F-Score, pp. 13; “The F-Score is a measurement of performance for the obtained neighborhood by comparing against the base neighborhood. To compute the F-score, we interpret the result of a k-NN query as a binary class, 1 for data points within the neighborhood and 0 for data points outside the neighborhood. The ground truth labels of the data points correspond to that obtained using a centralized trusted mediator solution.” The model in this article discloses the use of an automated verification system. This will approximate a value which is the difference between the output of a model and the ground truth.)
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” (Average F-score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” Using the data from the prediction and ground truth labels this system is able to determine a value denoting an accuracy value. This can be used to check how accurate the initial prediction is.)
“receiving, by the first computing system and from each second computing system of the one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” (Mediator Model, pp. 10; “In case of an untrusted mediator, the parties share either obfuscated data (randomized or encrypted data) or obfuscated local results with the mediator. The mediator has no knowledge of the data encryption scheme or the encryption keys used. As a result, even if the mediator is compromised, an adversary would not be able to decrypt the private data. However, if the adversary colludes with one or more parties, it might be able to partially reconstruct the data. We discuss these attacks in more detail later in this document.” One of the models proposed in this article can be a distributed or undistributed model with a untrusted mediator. This teaches that the individual users of the model cannot directly communicate raw data with the system or other users. This model is able to transmit data within the system using different forms of communications and protocols. This data transfer can include Average F-Score data which determines the accuracy of a labeled output.)
“determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;” (Average F-score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” Using the data from the prediction and ground truth labels this system is able to determine a value denoting an accuracy value. This can be used to check how accurate the initial prediction is. Equation 8 teaches the algorithm used to determine an accuracy value.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Perez and Montiel. Perez teaches a machine learning model which is able to determine a user based on indirect metadata. Montiel teaches a multiple party system which is able to preserve the privacy of its users and is able to communicate with user without revealing personal information. One of ordinary skill would have motivation to combine a machine learning model which is able to identify users based on their meta data and privacy-persevering multi-party environment where users are able to communicate without sharing personal information, "Figure 12 compares the privacy of the various algorithms. Note that the methods with Untrusted mediator offer a better privacy than those with trusted mediator. In particular, note that the CT method does not offer any privacy as it requires the parties to share the entire raw data with the mediator. On the other hand, the DT method offers some privacy as it only shares local results with the mediator. In contrast, the untrusted mediator methods offer different grades of privacy as a result of data obfuscation. In particular, we note that the SE encryption method (CU_se_) offers better privacy over the JLE encryption method (CU_jle_)." (Montiel, Overview of Results, pp. 21).
Regarding claim 2, Perez discloses, “wherein determining the predicted label for the given user profile comprises: determining, by the first computing system, a first share of the predicted label based at least in part on (i) the first share of the given user profile, (ii) the first machine learning model trained using the plurality of user profiles, and (iii) one or more of the plurality of true labels for the plurality of user profiles, the plurality of true labels including one or more true labels for each user profile in the plurality of user profiles;” (Dataset, pp. 4; “For data collection, we used the Twitter Streaming Public API (Twitter, Inc. 2018). Our population is a random1 sample of the tweets posted between October 2015 and January 2016 (inclusive). During this period we collected approximately 151,215,987 tweets corresponding 11,668,319 users. However, for the results presented here, we considered only users for which we collected more than 200 tweets. Our final dataset contains tweets generated by 5,412,693 users.” This system will determine a user based on non-identifying information. This model uses tweets from twitter and the users of twitter. This will make a determination using a machine learning model and is trained using ground truth labels.)
“receiving, by the first computing system and from the second computing system, data indicating a second share of the predicted label determined by the second computing system based at least in part on a second share of the given user profile and a first set of one or more machine learning models; and” (Dataset, pp. 4; “For data collection, we used the Twitter Streaming Public API (Twitter, Inc. 2018). Our population is a random1 sample of the tweets posted between October 2015 and January 2016 (inclusive). During this period we collected approximately 151,215,987 tweets corresponding 11,668,319 users. However, for the results presented here, we considered only users for which we collected more than 200 tweets. Our final dataset contains tweets generated by 5,412,693 users.” This system will determine a user based on non-identifying information. This model uses tweets from twitter and the users of twitter. This system is designed to execute using the internet and another instance of this model can be used to achieve a second result.)
“determining the predicted label based at least in part on the first and second shares of the predicted label.” (Attack Model, pp. 2; “We present the likelihood of success of an identification attack where the adversary’s ultimate goal is to identify a user from a set given this knowledge about the set of accounts. To achieve this, we answer this question: Is it possible to identify an individual from a set of metadata fields from a randomly selected set of Twitter user accounts?” This article teaches a method which is able to identify a user based on public meta data. This will return a result stating the inferred user for the given data. This teaches the output of a result from a machine learning model.)
Regarding claim 20, Perez discloses, “A system comprising: one or more processors of a first computing system; and one or more storage devices storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:” (Execution Time, pp. 7; “To compare the performance of the classifiers in terms of execution time we used a dedicated server with eight core Intel Xeon E5-2630 processors with 192GB DDR4 RAM running at 2,133MHz. For the implementation of the algorithms, we used Python 2.7 and Sci-kit learn release 0.17.1.” This method was executed and tested on a generic computing system as stated above using programming language. This system uses processors coupled with memory to perform actions in the machine language to produce a result.)
“receiving, by a first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” (Attack Model, pp. 2; “In our evaluation, as discussed before, the input of the classifier is a set of new (unseen) tweets. We refer to a successful prediction of the account identity as a hit and an unsuccessful one as a miss. We assume that the attacker is able to access the metadata of tweets from a group of users together with their identities (i.e., the training set) and that the new tweets belong to one of the users in the training set.” This model will take in user data, which does not directly identify the user, and is able to determine the user. This is used with twitter which teaches the use of a large social network containing multiple computing systems.)
“executing, by the first computing system in collaboration with one or more second computing systems of the plurality of MPC computing systems, a first machine learning model using the given user profile as input data, wherein the first machine learning model is trained using a plurality of user profiles;” (Implementation of the Classifiers, pp. 3; “The optimization of the internal parameters for each classifier was conducted as a combination of best practices in the field and experimental results in which we used the cross validated grid search capability offered by SciKit-learn. In summary, we calculated the value of the parameters for each classifier as follows. For KNN, we consider the single closest value based on the Euclidean distance between the observations; for RF, we chose entropy as the function to measure the effectiveness of the split of the data in a node; finally, for MLR, we selected the limited-memory implementation of the Broyden-Fletcher-Goldfarb-Shanno (LM-BFGS) optimizer as the value to optimize (Liu and Nocedal 1989).” The model in this article uses machine learning to predict a user based on user metadata. This metadata is public, which teaches the use of machine learning to identify a user based on outstanding, non-personal identifying data.)
“receiving, as an output of the first machine learning model, a predicted label for the given user profile;” (Attack Model, pp. 2; “We present the likelihood of success of an identification attack where the adversary’s ultimate goal is to identify a user from a set given this knowledge about the set of accounts. To achieve this, we answer this question: Is it possible to identify an individual from a set of metadata fields from a randomly selected set of Twitter user accounts?” This article teaches a method which is able to identify a user based on public meta data. This will return a result stating the inferred user for the given data. This teaches the output of a result from a machine learning model.)
“generating, by the first computing system, a first secret share of an inference result based at least in part on the predicted label determined for the given user profile and the predicted residue value;” (Implementation of the Classifiers, pp. 3; “The optimization of the internal parameters for each classifier was conducted as a combination of best practices in the field and experimental results in which we used the cross validated grid search capability offered by SciKit-learn. In summary, we calculated the value of the parameters for each classifier as follows. For KNN, we consider the single closest value based on the Euclidean distance between the observations; for RF, we chose entropy as the function to measure the effectiveness of the split of the data in a node; finally, for MLR, we selected the limited-memory implementation of the Broyden-Fletcher-Goldfarb-Shanno (LM-BFGS) optimizer as the value to optimize (Liu and Nocedal 1989).” The model in this article uses machine learning to predict a user based on user metadata. This metadata is public, which teaches the use of machine learning to identify a user based on outstanding, non-personal identifying data.)
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” (Dynamic Attributes, pp. 6; “Table 5 presents similar results for the RF algorithm. Even without the ACT, we are able to achieve 94.41% accuracy in a group of 10,000 users. These results are directly linked to the behavior of an account and are obtained from a multiclass model.” Tables 4-7 disclose the results of the experiments conducted by this model. This represents the returned values of how accurate the model was at identifying a user based on different levels of user attributes. This teaches the final results of the system and how accurate it was.)
Perez fails to explicitly disclose, “executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates a predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:”, “receiving, from each second computing system, an encrypted second secret share of the inference result;”, “executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;”, “receiving, by the first computing system and from each second computing system of the one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” and “determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;”.
However, Montiel discloses, “executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates a predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:” (Average F-Score, pp. 13; “The F-Score is a measurement of performance for the obtained neighborhood by comparing against the base neighborhood. To compute the F-score, we interpret the result of a k-NN query as a binary class, 1 for data points within the neighborhood and 0 for data points outside the neighborhood. The ground truth labels of the data points correspond to that obtained using a centralized trusted mediator solution.” The model in this article discloses the use of an automated verification system. This will approximate a value which is the difference between the output of a model and the ground truth.)
“receiving, from each second computing system, an encrypted second secret share of the inference result;” (Data Obfuscation Model, pp. 10; “As describe above, in case of an untrusted mediator, the parties further encrypt or obfuscate their data. We study two data obfuscation schemes in this work, Johnson-Lindenstrauss embedding and secure binary embedding. In both cases, the party’s will need to agree apriori on a common encryption key. The encryption key is secret to the parties, which will be used to decrypt the results back.” The model in this article discloses different models in a multiple party computation environment. Two models disclosed in the article described the use of untrusted mediators. A system using this proposed architecture uses encryption methods to transmit data from user to user or user to the mediator. Figure 6 shows this process of encryption and data transferring.)
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” (Average F-score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” Using the data from the prediction and ground truth labels this system is able to determine a value denoting an accuracy value. This can be used to check how accurate the initial prediction is.)
“receiving, by the first computing system and from each second computing system of the one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” (Mediator Model, pp. 10; “In case of an untrusted mediator, the parties share either obfuscated data (randomized or encrypted data) or obfuscated local results with the mediator. The mediator has no knowledge of the data encryption scheme or the encryption keys used. As a result, even if the mediator is compromised, an adversary would not be able to decrypt the private data. However, if the adversary colludes with one or more parties, it might be able to partially reconstruct the data. We discuss these attacks in more detail later in this document.” One of the models proposed in this article can be a distributed or undistributed model with a untrusted mediator. This teaches that the individual users of the model cannot directly communicate raw data with the system or other users. This model is able to transmit data within the system using different forms of communications and protocols. This data transfer can include Average F-Score data which determines the accuracy of a labeled output.)
“determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;” (Average F-score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” Using the data from the prediction and ground truth labels this system is able to determine a value denoting an accuracy value. This can be used to check how accurate the initial prediction is. Equation 8 teaches the algorithm used to determine an accuracy value.)
Regarding claim 21, Perez discloses, “A non-transitory computer readable storage medium carrying instructions that, when executed by one or more processors of a first computing system, cause the one or more processors to perform operations comprising:” Perez (Execution Time, pp. 7; “To compare the performance of the classifiers in terms of execution time we used a dedicated server with eight core Intel Xeon E5-2630 processors with 192GB DDR4 RAM running at 2,133MHz. For the implementation of the algorithms, we used Python 2.7 and Sci-kit learn release 0.17.1.” This method was executed and tested on a generic computing system as stated above using programming language. This system uses processors coupled with memory to perform actions in the machine language to produce a result.)
“receiving, by the first computing system of a plurality of multi-party computation (MPC) computing systems, an inference request comprising a first secret share of a given user profile;” (Attack Model, pp. 2; “In our evaluation, as discussed before, the input of the classifier is a set of new (unseen) tweets. We refer to a successful prediction of the account identity as a hit and an unsuccessful one as a miss. We assume that the attacker is able to access the metadata of tweets from a group of users together with their identities (i.e., the training set) and that the new tweets belong to one of the users in the training set.” This model will take in user data, which does not directly identify the user, and is able to determine the user. This is used with twitter which teaches the use of a large social network containing multiple computing systems.)
“executing, by the first computing system in collaboration with one or more second computing systems of the plurality of MPC computing systems, a first machine learning model using the given user profile as input data, wherein the first machine learning model is trained using a plurality of user profiles;” (Implementation of the Classifiers, pp. 3; “The optimization of the internal parameters for each classifier was conducted as a combination of best practices in the field and experimental results in which we used the cross validated grid search capability offered by SciKit-learn. In summary, we calculated the value of the parameters for each classifier as follows. For KNN, we consider the single closest value based on the Euclidean distance between the observations; for RF, we chose entropy as the function to measure the effectiveness of the split of the data in a node; finally, for MLR, we selected the limited-memory implementation of the Broyden-Fletcher-Goldfarb-Shanno (LM-BFGS) optimizer as the value to optimize (Liu and Nocedal 1989).” The model in this article uses machine learning to predict a user based on user metadata. This metadata is public, which teaches the use of machine learning to identify a user based on outstanding, non-personal identifying data.)
“receiving, as an output of the first machine learning model, a predicted label for the given user profile;” (Attack Model, pp. 2; “We present the likelihood of success of an identification attack where the adversary’s ultimate goal is to identify a user from a set given this knowledge about the set of accounts. To achieve this, we answer this question: Is it possible to identify an individual from a set of metadata fields from a randomly selected set of Twitter user accounts?” This article teaches a method which is able to identify a user based on public meta data. This will return a result stating the inferred user for the given data. This teaches the output of a result from a machine learning model.)
“generating, by the first computing system, a first secret share of an inference result based at least in part on the predicted label determined for the given user profile and the predicted residue value;” (Implementation of the Classifiers, pp. 3; “The optimization of the internal parameters for each classifier was conducted as a combination of best practices in the field and experimental results in which we used the cross validated grid search capability offered by SciKit-learn. In summary, we calculated the value of the parameters for each classifier as follows. For KNN, we consider the single closest value based on the Euclidean distance between the observations; for RF, we chose entropy as the function to measure the effectiveness of the split of the data in a node; finally, for MLR, we selected the limited-memory implementation of the Broyden-Fletcher-Goldfarb-Shanno (LM-BFGS) optimizer as the value to optimize (Liu and Nocedal 1989).” The model in this article uses machine learning to predict a user based on user metadata. This metadata is public, which teaches the use of machine learning to identify a user based on outstanding, non-personal identifying data.)
“providing, by the first computing system and to a client device, the first secret share of the inference result and each encrypted second secret share of the inference result received from the second computing system.” (Dynamic Attributes, pp. 6; “Table 5 presents similar results for the RF algorithm. Even without the ACT, we are able to achieve 94.41% accuracy in a group of 10,000 users. These results are directly linked to the behavior of an account and are obtained from a multiclass model.” Tables 4-7 disclose the results of the experiments conducted by this model. This represents the returned values of how accurate the model was at identifying a user based on different levels of user attributes. This teaches the final results of the system and how accurate it was.)
Perez fails to explicitly disclose, “executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:”, “receiving, from each second computing system, an encrypted second secret share of the inference result; and”, “executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising at the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;”, “receiving, by the first computing system and from [[a]] each second computing system of the plurality of 1VIPC computing systems one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” and “determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;”.
However, Montiel discloses, “executing, by the first computing system, a secure MPC protocol in collaboration with the one or more second computing systems to determine a predicted residue value for the given user profile, wherein the predicted residue value indicates predicted error between the predicted label for the given user profile and a true label for the given user profile, comprising:” (Average F-Score, pp. 13; “The F-Score is a measurement of performance for the obtained neighborhood by comparing against the base neighborhood. To compute the F-score, we interpret the result of a k-NN query as a binary class, 1 for data points within the neighborhood and 0 for data points outside the neighborhood. The ground truth labels of the data points correspond to that obtained using a centralized trusted mediator solution.” The model in this article discloses the use of an automated verification system. This will approximate a value which is the difference between the output of a model and the ground truth.)
“receiving, from each second computing system, an encrypted second secret share of the inference result; and” (Data Obfuscation Model, pp. 10; “As describe above, in case of an untrusted mediator, the parties further encrypt or obfuscate their data. We study two data obfuscation schemes in this work, Johnson-Lindenstrauss embedding and secure binary embedding. In both cases, the party’s will need to agree apriori on a common encryption key. The encryption key is secret to the parties, which will be used to decrypt the results back.” The model in this article discloses different models in a multiple party computation environment. Two models disclosed in the article described the use of untrusted mediators. A system using this proposed architecture uses encryption methods to transmit data from user to user or user to the mediator. Figure 6 shows this process of encryption and data transferring.)
“executing, by the first computing system in collaboration with the one or more second computing systems, a second machine learning model to obtain a first secret share of the predicted residue value for the given user profile based on an input comprising at the first secret share of the given user profile, wherein the second machine learning model is trained using the plurality of user profiles and residue value data indicating differences between a plurality of true labels for the plurality of user profiles and a plurality of predicted labels as determined for the plurality of user profiles using the first machine learning model, wherein the second machine learning model is different from the first machine learning model, and wherein the first computing system and each of the one or more second computing systems perform computations of both the first and second machine learning models using secret shares of data and the secure MPC protocol;” (Average F-score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” Using the data from the prediction and ground truth labels this system is able to determine a value denoting an accuracy value. This can be used to check how accurate the initial prediction is.)
“receiving, by the first computing system and from [[a]] each second computing system of the plurality of 1VIPC computing systems one or more second computing systems, data indicating a respective second secret share of the predicted residue value for the given user profile determined by the second computing system based at least in part on respective second secret share of the given user profile and a second set of one or more machine learning models; and” (Mediator Model, pp. 10; “In case of an untrusted mediator, the parties share either obfuscated data (randomized or encrypted data) or obfuscated local results with the mediator. The mediator has no knowledge of the data encryption scheme or the encryption keys used. As a result, even if the mediator is compromised, an adversary would not be able to decrypt the private data. However, if the adversary colludes with one or more parties, it might be able to partially reconstruct the data. We discuss these attacks in more detail later in this document.” One of the models proposed in this article can be a distributed or undistributed model with a untrusted mediator. This teaches that the individual users of the model cannot directly communicate raw data with the system or other users. This model is able to transmit data within the system using different forms of communications and protocols. This data transfer can include Average F-Score data which determines the accuracy of a labeled output.)
“determining the predicted residue value for the given user profile based at least in part on the first and second secret shares of the predicted residue value;” (Average F-score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” Using the data from the prediction and ground truth labels this system is able to determine a value denoting an accuracy value. This can be used to check how accurate the initial prediction is. Equation 8 teaches the algorithm used to determine an accuracy value.)
Claims 3-5 are rejected under 35 U.S.C. 103 as being unpatentable over Perez and Montiel in view of Kenthapadi et al., (Kenthapadi et al., “Privacy via the Johnson-Lindenstrauss Transform”, April 2012, hereinafter “Kenthapadi”).
Regarding claim 3, Perez discloses, “determining, by the first computing system, a first share of the predicted label based at least in part on the first transformed share of the given user profile.” (Obfuscation and Re-Identification, pp. 4; “We measured the level of protection awarded by randomization by recording the accuracy of the predictions as we increased the number of obfuscated data points in increments of 10% until we reached full anonymization (i.e., 100% randomization) of the training set.” The model in this article is able to use encrypted data and is still able to make a prediction. This teaches that the proposed system is able to take in transformed data and perform operation on that data. Then this model is able to produce a final result using that transformed data.)
Perez and Montiel fail to explicitly disclose, “applying, by the first computing system, a transformation to the first share of the given user profile to obtain a first transformed share of the given user profile, wherein determining, by the first computing system, the first share of the predicted label comprises:”.
However, Kenthapadi discloses, “applying, by the first computing system, a transformation to the first share of the given user profile to obtain a first transformed share of the given user profile, wherein determining, by the first computing system, the first share of the predicted label comprises:” (User representation, pp. 3; "We represent each user belonging to a set U of n users as a binary vector in d dimensions, where each dimension corresponds to the value of an attribute (e.g. gender, interest/disinterest in a particular topic, location information, etc.) We assume that the attribute meanings are not sensitive or, if they are, they can be published in a privacy-preserving manner (say, using the techniques in Goetz et al. or Korolova et al. [16, 20]). Our goal can be formally stated as: Given a set of user profiles represented as vectors in d dimensions, publish sketches of the user profiles that simultaneously preserve user privacy and enable third parties to estimate pairwise distance between users." The data used in this case can be transformed in order to maintain it privacy between users in a system. This article describes how user data can be handled by multiple users with keeping the original user’s data private using transformations.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Perez, Montiel and Kenthapadi. Perez teaches a machine learning model which is able to determine a user based on indirect metadata. Montiel teaches a multiple party system which is able to preserve the privacy of its users and is able to communicate with user without revealing personal information. Kenthapadi teaches a method that transforms user data in order to hide and protect that data from outside sources. One of ordinary skill would have motivation to combine a machine learning model which is able to identify users based on their meta data with a privacy-persevering multi-party environment where users are able to communicate without sharing personal information and with a system that is able to transform data to add a layer of security, "The key insight behind our technique is that by projecting users to a lower-dimensional space, we can limit the amount of noise we add to each user's data, while also reaping the benefit of preserving distances. We also compared our proposed solution to other candidate solutions, such as directly adding noise to the pairwise distances or adding noise to each attribute of a user, and showed that our method is preferable for potential applications such as user segmentation and nearest neighbor search." (Kenthapadi, Conclusion and Future Work, pp. 21).
Regarding claim 4, Perez and Montiel fail to explicitly disclose, “wherein the transformation comprises a Johnson-Lindenstrauss (J-L) transformation.”.
However, Kenthapadi discloses, “wherein the transformation comprises a Johnson-Lindenstrauss (J-L) transformation.” (Abstract, pp. 1; "Our method involves projecting each user's representation into a random, lower-dimensional space via a sparse Johnson- Lindenstrauss transform and then adding Gaussian noise to each entry of the lower dimensional representation. We show that the method preserves differential privacy-where the more privacy is desired, the larger the variance of the Gaussian noise." This article discloses a method of transforming data using the Johnson-Lindenstrauss (J-L) transformation.)
Regarding claim 5, Perez discloses, “wherein determining, by the first computing system, the first share of the predicted label comprises: providing, by the first computing system, the first transformed share of the given user profile as input to the first machine learning model to obtain a first share of the predicted label for the given user profile as output.” (Attack Model, pp. 2; “We present the likelihood of success of an identification attack where the adversary’s ultimate goal is to identify a user from a set given this knowledge about the set of accounts. To achieve this, we answer this question: Is it possible to identify an individual from a set of metadata fields from a randomly selected set of Twitter user accounts?” This article teaches a method which is able to identify a user based on public meta data. This will return a result stating the inferred user for the given data. This teaches the output of a result from a machine learning model.)
Claims 6 and 11-15 are rejected under 35 U.S.C. 103 as being unpatentable over Perez and Montiel in view of Zhao et al., (Zhao et al., “Privacy-Preserving Collaborative Deep Learning With Unreliable Participants”, 2020, hereinafter “Zhao”).
Regarding claim 6, Perez discloses, “evaluating a performance of the first machine learning model, comprising, for each of the plurality of user profiles: determining a predicted label for the user profile, comprising: determining, by the first computing system, a first share of a predicted label for the user profile based at least in part on (i) a first share of the user profile, (ii) the first machine learning model, and (iii) one or more of the plurality of true labels for the plurality of user profiles;” (Dataset, pp. 4; “For data collection, we used the Twitter Streaming Public API (Twitter, Inc. 2018). Our population is a random1 sample of the tweets posted between October 2015 and January 2016 (inclusive). During this period we collected approximately 151,215,987 tweets corresponding 11,668,319 users. However, for the results presented here, we considered only users for which we collected more than 200 tweets. Our final dataset contains tweets generated by 5,412,693 users.” This system will determine a user based on non-identifying information. This model uses tweets from twitter and the users of twitter. This will make a determination using a machine learning model and is trained using ground truth labels.)
“receiving, by the first computing system and from the second computing system, data indicating a second share of the predicted label for the user profile determined by the second computing system based at least in part on a second share of the user profile and the first set of one or more machine learning models maintained by the second computing system; and” (Dataset, pp. 4; “For data collection, we used the Twitter Streaming Public API (Twitter, Inc. 2018). Our population is a random1 sample of the tweets posted between October 2015 and January 2016 (inclusive). During this period we collected approximately 151,215,987 tweets corresponding 11,668,319 users. However, for the results presented here, we considered only users for which we collected more than 200 tweets. Our final dataset contains tweets generated by 5,412,693 users.” This system will determine a user based on non-identifying information. This model uses tweets from twitter and the users of twitter. This system is designed to execute using the internet and another instance of this model can be used to achieve a second result.)
“determining the predicted label for the user profile based at least in part on the first and second shares of the predicted label;” (Attack Model, pp. 2; “We present the likelihood of success of an identification attack where the adversary’s ultimate goal is to identify a user from a set given this knowledge about the set of accounts. To achieve this, we answer this question: Is it possible to identify an individual from a set of metadata fields from a randomly selected set of Twitter user accounts?” This article teaches a method which is able to identify a user based on public meta data. This will return a result stating the inferred user for the given data. This teaches the output of a result from a machine learning model.)
Perez fails to explicitly disclose, “determining a residue value for the user profile indicating an error in the predicted label, comprising: determining, by the first computing system, a first share of the residue value for the user profile based at least in part on the predicted label determined for the user profile and a first share of a true label for the user profile included in the plurality of true labels;”, “receiving, by the first computing system and from the second computing system, data indicating a second share of the residue value for the user profile determined by the second computing system based at least in part on the predicted label determined for the user profile and a second share of the true label for the user profile; and”, “determining the residue value for the user profile based at least in part on the first and second shares of the residue value; and” and “training the second machine learning model using data indicating the residue values determined for the plurality of user profiles in evaluating the performance of the first machine learning model.”.
However, Montiel discloses, “determining a residue value for the user profile indicating an error in the predicted label, comprising: determining, by the first computing system, a first share of the residue value for the user profile based at least in part on the predicted label determined for the user profile and a first share of a true label for the user profile included in the plurality of true labels;” (Average F-score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” Using the data from the prediction and ground truth labels this system is able to determine a value denoting an accuracy value. This can be used to check how accurate the initial prediction is.)
“receiving, by the first computing system and from the second computing system, data indicating a second share of the residue value for the user profile determined by the second computing system based at least in part on the predicted label determined for the user profile and a second share of the true label for the user profile; and” Mediator Model (Montiel, pp. 10; “In case of an untrusted mediator, the parties share either obfuscated data (randomized or encrypted data) or obfuscated local results with the mediator. The mediator has no knowledge of the data encryption scheme or the encryption keys used. As a result, even if the mediator is compromised, an adversary would not be able to decrypt the private data. However, if the adversary colludes with one or more parties, it might be able to partially reconstruct the data. We discuss these attacks in more detail later in this document.” One of the models proposed in this article can be a distributed or undistributed model with a untrusted mediator. This teaches that the individual users of the model cannot directly communicate raw data with the system or other users. This model is able to transmit data within the system using different forms of communications and protocols. This data transfer can include Average F-Score data which determines the accuracy of a labeled output.)
“determining the residue value for the user profile based at least in part on the first and second shares of the residue value; and” (Average F-score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” Using the data from the prediction and ground truth labels this system is able to determine a value denoting an accuracy value. This can be used to check how accurate the initial prediction is. Equation 8 teaches the algorithm used to determine an accuracy value.)
Perez and Montiel fail to explicitly disclose, “training the second machine learning model using data indicating the residue values determined for the plurality of user profiles in evaluating the performance of the first machine learning model.”.
However, Zhao discloses, “training the second machine learning model using data indicating the residue values determined for the plurality of user profiles in evaluating the performance of the first machine learning model.” (SecProbe: The Participant Part, pp. 1492" "Algorithm 3 presents the pseudocode of SecProbe on the participant side. Each participant has its own local training dataset and conducts the standard SGD algorithm to train its local model. Let
W
i
denote the network weights of participant i. To protect the privacy of the participant's sensitive data being disclosed by
W
i
, the participant applies differential privacy onto the training algorithm to get sanitized weights
W
i
-
'
s
, and uploads them to the server." This article discloses a second machine learning model which is trained locally. This model is train from the values of the training data and will be taught to mimic the global model for security.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Perez, Montiel and Zhao. Perez teaches a machine learning model which is able to determine a user based on indirect metadata. Montiel teaches a multiple party system which is able to preserve the privacy of its users and is able to communicate with user without revealing personal information. Zhao teaches a method which uses privacy-preserving techniques and deep learning to generate collaborative user networks. One of ordinary skill would have motivation to combine a machine learning model which is able to identify users based on their meta data with a privacy-persevering multi-party environment where users are able to communicate without sharing personal information and a system which uses deep learning for assistance with security to build a collaborative user environment while also preserving user privacy, "SecProbe utilizes exponential mechanism and functional mechanism to protect both the privacy of the participants' data and the quality of their data, the two major privacy concerns in such a system. The experimental results demonstrated that SecProbe is robust to unreliable participants, and can achieve high-accuracy results which are close to the model trained in a traditional centralized manner, while providing rigorous privacy guarantee." (Zhao, Conclusion, pp. 1498).
Regarding claim 11, Perez fails to explicitly disclose, “the first machine learning model includes a k-nearest neighbor model maintained by the first computing system;”, “the first set of one or more machine learning models includes a k-nearest neighbor model maintained by the second computing system;”, “the second machine learning model includes at least one of a deep neural network (DNN) maintained by the first computing system and a gradient-boosting decision tree (GBDT) maintained by the first computing system; and”, and “the second set of one or more machine learning models includes at least one of a DNN maintained by the second computing system and a GBDT maintained by the second computing system.”.
However, Montiel discloses, “the first machine learning model includes a k-nearest neighbor model maintained by the first computing system;” (Computational Model, pp. 8; "In case of a distributed computation model, each party builds a local model based on their local data. During the query phase, a party will send its query to the mediator. The mediator shares the query with all participating parties. Each party will then execute a k-NN search on their local data, and share the local results with the mediator. Finally, the mediator aggregates and filters the local results to compute the global k-NN results, and shares them with the query originator." This article discloses the use of a k-NN model use in a multiparty computation network for social networking. This is used to find pair or like cases in users and data.)
“the first set of one or more machine learning models includes a k-nearest neighbor model maintained by the second computing system;” (Computational Model, pp. 8; "In case of a distributed computation model, each party builds a local model based on their local data. During the query phase, a party will send its query to the mediator. The mediator shares the query with all participating parties. Each party will then execute a k-NN search on their local data, and share the local results with the mediator. Finally, the mediator aggregates and filters the local results to compute the global k-NN results, and shares them with the query originator." This article discloses the use of a k-NN model use in a multiparty computation network for social networking. This is used to find pair or like cases in users and data. The k-NN model is located on the local device and can be performed by any device in the network.) Perez and Montiel fail to explicitly disclose, “the second machine learning model includes at least one of a deep neural network (DNN) maintained by the first computing system and a gradient-boosting decision tree (GBDT) maintained by the first computing system; and”, and “the second set of one or more machine learning models includes at least one of a DNN maintained by the second computing system and a GBDT maintained by the second computing system.”.
However, Zhao discloses, “the second machine learning model includes at least one of a deep neural network (DNN) maintained by the first computing system and a gradient-boosting decision tree (GBDT) maintained by the first computing system; and” (SecProbe: The Participant Part, pp. 1492; "Since the structures of the neural networks may be varied and often depend on specific application scenarios, it is impossible to design a one-size-fits-all differentially-private solution for all deep learning models. In this paper, we focus on the most common neural network MLP. Specifically, we first consider a three-layer fully-connected neural network, design algorithms to train the model in a differentially-private manner, and then show that more hidden layers can be stacked easily by using our proposed scheme." This article discloses a system that uses locally trained models. These models are described earlier in the article as Deep Learning Neural Networks which utilize a multi-layer Model using Re LU activation function to help improve accuracy.)
“the second set of one or more machine learning models includes at least one of a DNN maintained by the second computing system and a GBDT maintained by the second computing system.” (SecProbe: The Participant Part, pp. 1492; "Since the structures of the neural networks may be varied and often depend on specific application scenarios, it is impossible to design a one-size-fits-all differentially-private solution for all deep learning models. In this paper, we focus on the most common neural network MLP. Specifically, we first consider a three-layer fully-connected neural network, design algorithms to train the model in a differentially-private manner, and then show that more hidden layers can be stacked easily by using our proposed scheme." This article discloses a system that uses locally trained models. These models are described earlier in the article as Deep Learning Neural Networks which utilize a multi-layer Model using Re LU activation function to help improve accuracy.)
Regarding claim 12, Perez and Zhao fail to explicitly disclose, “wherein determining, by the first computing system, the first share of the predicted label comprises: identifying, by the first computing system, a first set of nearest neighbor user profiles based at least in part on the first share of the given user profile and the k-nearest neighbor model maintained by the first computing system;”, “receiving, by the first computing system and from the second computing system, data indicating a second set of nearest neighbor profiles identified by the second computing system based at least in part on the second share of the given user profile and the k-nearest neighbor model maintained by the second computing system;”, “identifying a number k of nearest neighbor user profiles that are considered most similar to the given user profile among the plurality of user profiles based at least in part on the first and second sets of nearest neighbor profiles; and” and “determining, by the first computing system, the first share of the predicted label based at least in part on a true label for each of the k nearest neighbor user profiles.”
However, Montiel discloses, “wherein determining, by the first computing system, the first share of the predicted label comprises: identifying, by the first computing system, a first set of nearest neighbor user profiles based at least in part on the first share of the given user profile and the k-nearest neighbor model maintained by the first computing system;” (Average Neighbor Distance, pp. 14-15; "This is the simplest approach of accuracy measurement which computes the mean distance of all neighbors against the query data point. The average neighbor distance measure does not require the knowledge of ground truth neighbors, and hence is easy to compute. However, it is not a normalized metric and should be quantified relatively. Given a query and its k neighbors, the average neighbor distance is computed as, [See Equation 6].” This figure shows an example of how to locate the average distance of neighbors in the system. To find the average the distance off all the neighbors is needed.)
“receiving, by the first computing system and from the second computing system, data indicating a second set of nearest neighbor profiles identified by the second computing system based at least in part on the second share of the given user profile and the k-nearest neighbor model maintained by the second computing system;” (Average Neighbor Distance, pp. 14-15; "This is the simplest approach of accuracy measurement which computes the mean distance of all neighbors against the query data point. The average neighbor distance measure does not require the knowledge of ground truth neighbors, and hence is easy to compute. However, it is not a normalized metric and should be quantified relatively. Given a query and its k neighbors, the average neighbor distance is computed as, [See Equation 6].” This figure shows an example of how to locate the average distance of neighbors in the system. To find the average the distance off all the neighbors is needed. Since this can be completed on a distributed model than this can be performed on multiple different devices in the network.)
“identifying a number k of nearest neighbor user profiles that are considered most similar to the given user profile among the plurality of user profiles based at least in part on the first and second sets of nearest neighbor profiles; and” (Privacy-preserving k-NN Computation, pp. 8; "In this section, we study the problem of privacy-preserving k-NN, where the goal is to identify the k nearest neighbors from distributed multi-party data, while preserving the data privacy of each individual party. As an example, consider a healthcare scenario, where a group of hospitals wish to collaborate and work together to improve the collective quality of healthcare. Each hospital already collects a lot of data about its patients, including their demographics, past medical history, lab results, current diagnosis, prescribed treatment and outcomes. This data contains a wealth of information that if shared across the group could mutually benefit all parties by enabling faster diagnosis and effective treatment for similar cases." The main goal of this article is to identify and determine similarities in data by preserving the privacy of the user. This will use k-NN to help determine the similarities in the data and users.)
“determining, by the first computing system, the first share of the predicted label based at least in part on a true label for each of the k nearest neighbor user profiles.” (Privacy-preserving k-NN Computation, pp. 8; "In this section, we study the problem of privacy preserving k-NN, where the goal is to identify the k nearest neighbors from distributed multiparty data, while preserving the data privacy of each individual party. As an example, consider a healthcare scenario, where a group of hospitals wish to collaborate and work together to improve the collective quality of healthcare. Each hospital already collects a lot of data about its patients, including their demographics, past medical history, lab results, current diagnosis, prescribed treatment and outcomes. This data contains a wealth of information that if shared across the group could mutually benefit all parties by enabling faster diagnosis and effective treatment for similar cases." The main goal of this article is to identify and determine similarities in data by preserving the privacy of the user. This will use k-NN to help determine the similarities in the data and users.)
Regarding claim 13, Perez and Zhao fail to explicitly disclose, “wherein determining, by the first computing system, the first share of the predicted label further comprises: determining, by the first computing system, a first share of a sum of the true labels for the k nearest neighbor user profiles;”, “receiving, by the first computing system and from the second computing system, a second share of the sum of the true labels for the k nearest neighbor user profiles; and”, and “determining the sum of the true labels for the k nearest neighbor user profiles based at least in part on the first and second shares of the sum of the true labels for the k nearest neighbor user profiles.”
However, Montiel discloses, “wherein determining, by the first computing system, the first share of the predicted label further comprises: determining, by the first computing system, a first share of a sum of the true labels for the k nearest neighbor user profiles;” (Average F-Score, pp. 13-14; "The F-Score is a measurement of performance for the obtained neighborhood by comparing against the base neighborhood. To compute the F-score, we interpret the result of a k-NN query as a binary class, 1 for data points within the neighborhood and 0 for data points outside the neighborhood. The ground truth labels of the data points correspond to that obtained using a centralized trusted mediator solution." In this article the NN are evaluated. This evaluation will compute the F-score which is the number of datapoint within a neighbor and can use ground truths to help determine this in some models.)
“receiving, by the first computing system and from the second computing system, a second share of the sum of the true labels for the k nearest neighbor user profiles; and” (Average F-Score, pp. 13-14; "The F-Score is a measurement of performance for the obtained neighborhood by comparing against the base neighborhood. To compute the F-score, we interpret the result of a k-NN query as a binary class, 1 for data points within the neighborhood and 0 for data points outside the neighborhood. The ground truth labels of the data points correspond to that obtained using a centralized trusted mediator solution." In this article the NN are evaluated. This evaluation will compute the F-score which is the number of datapoint within a neighbor and can use ground truths to help determine this in some models. This is performed on a local device depending on the model used in this article. This teaches that a second computing system can perform this action and send its results to another party in the network)
“determining the sum of the true labels for the k nearest neighbor user profiles based at least in part on the first and second shares of the sum of the true labels for the k nearest neighbor user profiles.” (Average F-Score, pp. 14; "Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score." To calculate the F-score, the scores of different devices can be used to get an overall result.)
Regarding claim 14, Perez and Zhao fail to explicitly disclose, “wherein determining, by the first computing system, the first share of the predicted label further comprises: applying a function to the sum of the true labels for the k nearest neighbor user profiles to generate the first share of the predicted label for the given user profile.”.
However, Montiel discloses, “wherein determining, by the first computing system, the first share of the predicted label further comprises: applying a function to the sum of the true labels for the k nearest neighbor user profiles to generate the first share of the predicted label for the given user profile.” (Average F-Score, pp. 14; "Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score." To calculate the F-score, the scores of different devices can be used to get an overall result. This F-Score could be transmitted to other system in the network for various uses and to help determine accuracy of the model.)
Regarding claim 15, Perez and Zhao fail to explicitly disclose, “wherein the first share of the predicted label for the given user profile comprises the sum of the true labels for the k nearest neighbor user profiles.”
However, Montiel discloses, “wherein the first share of the predicted label for the given user profile comprises the sum of the true labels for the k nearest neighbor user profiles.” (Average F-Score, pp. 14; "Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score." To calculate the F-score, the scores of different devices can be used to get an overall result. This F-Score could be transmitted to other system in the network for various uses and to help determine the accuracy of the model.)
Claims 7-10 are rejected under 35 U.S.C. 103 as being unpatentable over Perez, Montiel and Zhao in view of Shukla et al., (Shukla et al., “SYSTEMIS AND METHODS FOR VERIFIABLE, PRIVATE, AND SECURE OMIC ANALYSIS”, US 2015/0213079 A1, filed Jan. 2015, hereinafter “Shukla”).
Regarding claim 7, Perez discloses, “configuring the first machine learning model to, given a user profile as input, generate an initial predicted label for the user profile and apply the function, as defined based on the derived set of parameters, to the initial predicted label for the user profile to generate, as output, a first share of a predicted label for the user profile.” (Attack Model, pp. 2; “We present the likelihood of success of an identification attack where the adversary’s ultimate goal is to identify a user from a set given this knowledge about the set of accounts. To achieve this, we answer this question: Is it possible to identify an individual from a set of metadata fields from a randomly selected set of Twitter user accounts?” The model in this article discloses the use of a machine learning model to take in user data and make a to determine a user based on input data. This model is trained on twitter meta in order to produce an output.)
Perez, Montiel and Zhao fail to explicitly disclose, “before evaluating the performance of the first machine learning model: deriving a set of parameters of a function, comprising: deriving, by the first computing system, a first share of the set of parameters of the function based at least in part on a first share of each of the plurality of true labels;”, “receiving, by the first computing system and from the second computing system, data indicating a second share of the set of parameters of the function derived by the second computing system based at least in part on a second share of each of the plurality of true labels; and” and “deriving the set of parameters of the function based at least in part on the first and second shares of the set of parameters of the function; and”.
However, Shukla discloses, “before evaluating the performance of the first machine learning model: deriving a set of parameters of a function, comprising: deriving, by the first computing system, a first share of the set of parameters of the function based at least in part on a first share of each of the plurality of true labels;” (Verifiable, Private and Secure Omic Matching, pp. 19, [0131]; “Continuing the example, the mean and variance of the posterior distribution may be calculated for particular values of the parameters, and the mean or the mode (maximum likelihood estimate) of the posterior distribution may be used as an estimate of t; estimates of t and its associated variance for different parameter values may also be precomputed and stored in a lookup table. A set of parameters are calculated in this invention to be used for determining relations between users. The set of values are can be calculated by a first computing device.)
“receiving, by the first computing system and from the second computing system, data indicating a second share of the set of parameters of the function derived by the second computing system based at least in part on a second share of each of the plurality of true labels; and” (Verifiable, Private and Secure Omic Matching, pp. 19, [0132]; “Continuing the example, the circuit may first calculate k by comparing (a,b,) pairs and then look up the above table and fetch the appropriate value for t and its associated standard deviation, and then use a counter to output the number of matches k; in this case (n=5), the output is 3 bits long.” In this article the set of parameters can be determine for multiple devices to find similarities. This discloses that a second computing device is able to use the same set of parameters.)
“deriving the set of parameters of the function based at least in part on the first and second shares of the set of parameters of the function; and” (Verifiable, Private and Secure Omic Matching, pp. 19, [0132]; “FIG. 8 depicts one embodiment of a circuit performing the calculation in the above example. In other embodiments, the circuit may incorporate mutation rates differing from the Infinite Alleles Model, such as the Stepwise Mutation Model.” Figure 8 discloses the combination of the two users derived data.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Perez, Montiel, Zhao and Shukla. Perez teaches a machine learning model which is able to determine a user based on indirect metadata. Montiel teaches a multiple party system which is able to preserve the privacy of its users and is able to communicate with user without revealing personal information. Zhao teaches a method which uses privacy-preserving techniques and deep learning to generate collaborative user networks. Shukla teaches a system that uses privacy preserving techniques with multiparty computation to store, manipulate and export user profile data. One of ordinary skill would have motivation to combine a machine learning model which is able to identify users based on their meta data with a privacy-persevering multi-party environment where users are able to communicate without sharing personal information with a system which uses deep learning for assistance with security to build a collaborative user environment while also preserving user privacy and a system that uses multi-party computation and automation to evaluate and transmit user data while preserving user privacy, (Shukla, Verifiable, Private and Secure Omic Matching, pp. 19, [0136]; “Embodiments of the above-described system and method permit users to discover medical issues latent in their omic profiles without fear of Stigma due to privacy breaches. A person who has a particular concern with regard to his or her omic profile may tailor the calculation to address that concern. In some embodiments, the use of secure multiparty calculation to discover the matching score ensures that neither party to the calculation discovers the omic data of the other party; the secure multiparty calculation also protects the privacy of the omic data from third-party listeners, including an authentication server. The use of secure multiparty computation and homomorphic encryption to Verify omic data may enable the authentication server to guarantee both that the omic data is genuine and that the authentication server retains no unencrypted version of the omic data, limiting fears of security breaches and data mining. Safeguards against malicious users help to ensure the privacy and authenticity of the calculations.”
Regarding claim 8, Perez, Montiel and Zhao fail to explicitly disclose, “estimating, by the first computing system, a first share of a set of distribution parameters based at least in part on the first share of each of the plurality of true labels, wherein deriving, by the first computing system, the first share of the set of parameters of the function based at least in part on the first share of each of the plurality of true labels comprises: deriving, by the first computing system, a first share of the set of parameters of the function based at least in part on the first share of the set of distribution parameters.”.
However, Shukla discloses, “estimating, by the first computing system, a first share of a set of distribution parameters based at least in part on the first share of each of the plurality of true labels, wherein deriving, by the first computing system, the first share of the set of parameters of the function based at least in part on the first share of each of the plurality of true labels comprises: deriving, by the first computing system, a first share of the set of parameters of the function based at least in part on the first share of the set of distribution parameters.” (Verifiable, Private and Secure Omic Matching, pp. 19, [0131]; “Continuing the example, the mean and variance of the posterior distribution may be calculated for particular values of the parameters, and the mean or the mode (maximum likelihood estimate) of the posterior distribution may be used as an estimate of t; estimates of t and its associated variance for different parameter values may also be precomputed and stored in a lookup table. A set of parameters are calculated in this invention to be used for determining relations between users. The set of values are can be calculated by a first computing device.)
Regarding claim 9, Perez, Montiel and Shukla fail to explicitly disclose, “wherein the set of distribution parameters include one or more parameters of a probability distribution of prediction errors for true labels of a first value in the plurality of true labels, and one or more parameters of a probability distribution of prediction errors for true labels of a second value in the plurality of true labels, the second value being different from the first value.”
However, Zhao discloses, “wherein the set of distribution parameters include one or more parameters of a probability distribution of prediction errors for true labels of a first value in the plurality of true labels, and one or more parameters of a probability distribution of prediction errors for true labels of a second value in the plurality of true labels, the second value being different from the first value.” (SecProbe: the Server Part, pp. 1491; "The privacy of the participants' training data will be considered in the next step. As it is unfair to measure the performance of the participants' own data, a reliable dataset is required for evaluation, e.g., a validation dataset on the server. In our design, since the performance of the model is reflected by the prediction accuracy, we protect the prediction values on the validation dataset. Specifically, the server constructs a centralized virtual dataset which has the same number of records with the original validation dataset, and each record has M attributes which represent the prediction results (the clipped zi for regression, or the correctness for classification) on the M participants' models." A central database is kept to help train all of the local Machine Learning models. In this article a central server contains the databased involved in initially training all of the local models. This server contains different weights, labels and parameters.)
Regarding claim 10, Perez, Zhao and Shukla fail to explicitly disclose, “the first share of the residue value for the user profile is indicative of a difference in value between the predicted label determined for the user profile and the first share of the true label for the user profile; and” and “the second share of the residue value for the user profile is indicative of a difference in value between the predicted label determined for the user profile and the second share of the true label for the user profile.”
However, Montiel discloses, “the first share of the residue value for the user profile is indicative of a difference in value between the predicted label determined for the user profile and the first share of the true label for the user profile; and” (Average F-Score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” The model in this article uses an evaluation to determine the accuracy of the output. This uses equation 8 to make output a score, or value, indicating the accuracy of the output.)
“the second share of the residue value for the user profile is indicative of a difference in value between the predicted label determined for the user profile and the second share of the true label for the user profile.” (Average F-Score, pp. 14; “Given the observed labels and the ground truth labels, we can evaluate the confusion matrix consisting of true positives, true negatives, false positives and false negatives. These values can then be used to compute precision and recall as shown in Equation 7. Both values are used to compute the F-score.” The system proposed in the article uses Mult-party computations and different machine are able to perform the above evaluation of an output. This teaches that a second system is able to perform the actions above as well as an initial system.)
Claims 16 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Perez, Montiel and Zhao in view of Park et al., (Park et al., “Parallelly Running k-Nearest Neighbor Classification Over Semantically Secure Encrypted Data in Outsourced Environments”, 2020, hereinafter “Park”).
Regarding claim 16, Perez, Montiel and Zhao fail to explicitly disclose, “wherein determining, by the first computing system, the first share of the predicted label based at least in part on the true label for each of the k nearest neighbor user profiles comprises:”, “determining, by the first computing system, a first share of a set of predicted labels based at least in part on a set of true labels for each of the k nearest neighbor user profiles corresponding to a set of categories, respectively, comprising, for each category in the set:”, “determining a first share of a frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of a first value;”, “receiving, by the first computing system and from the second computing system, a second share of the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value; and” and “determining the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value based at least in part on the first and second shares of the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value.”
However, Park discloses, “wherein determining, by the first computing system, the first share of the predicted label based at least in part on the true label for each of the k nearest neighbor user profiles comprises:” (Overview of SKLE and SKSE, pp. 64621; "However, SkLE and SkSE are carried out in parallel for each data or in common for all data (for example, computing the number of largest data and comparing the number to parameter k are carried out in common for all data). Furthermore, SkLE (resp., SkSE) find k largest (resp., smallest) elements in only one execution without changing the value of the highest (resp., lowest) element to the lowest (resp., highest) value." Each of the neighbors in the system are located and labeled accordingly. In this article they are labeled to find the largest and smallest kin distance from the model.)
“determining, by the first computing system, a first share of a set of predicted labels based at least in part on a set of true labels for each of the k nearest neighbor user profiles corresponding to a set of categories, respectively, comprising, for each category in the set:” (Overview of SKLE and SKSE, pp. 64622; "As mentioned in Section II-A, DH has input ciphertexts and CSP has a decryption key. In SkLE (resp., SkSE), each element
e
i
has auxiliary data
K
i
,
P
i
, and
C
i
whose details are as follows.
K
i
indicates whether an element ei is one of k largest (resp., smallest) elements where
K
i
= 1 if an element
e
i
is one of k largest (resp., smallest) elements and
K
i
= 0 otherwise. Once an element
e
i
becomes one of k largest (resp., smallest) elements, it is irreversible. (i.e.,
K
i
= 0
→
1 but 1
↛
0).
P
i
indicates whether an element
e
i
is a predicted largest (resp., smallest) element which means the element predicted as a largest (resp., smallest) element where
P
i
= 1 if an element
e
i
is a predicted largest (resp., smallest) element and
P
i
= 0 otherwise." The model proposed in this article uses the data given to label the neighbors data based on distance. This model will take note of all nearest and farthest neighbors by distance.)
“determining a first share of a frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of a first value;” (Algorithm 1, pp. 64622; This algorithm teaches how the model will discover neighbors and rank them by distance. This algorithm is able to count the number of neighbors and denotes by then number of neighbors.)
“receiving, by the first computing system and from the second computing system, a second share of the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value; and” (Algorithm 1, pp. 64622; This algorithm is designed to run on a local device and the results a can be output to other devices in the network. Other devices run this algorithm separately as well to get their own results.)
“determining the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value based at least in part on the first and second shares of the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value.” (Secure Comparison (SCP), pp. 64625; "We defined our comparison protocol (SCP) to compare the number of predicated largest or smallest elements (E(S)) with public parameter kin Section IV-B. In this subsection, we present SCP in Algorithm 3, which is designed by applying the idea in [23]." After the Algorithm 1 and 2 are executed on local devices the results are compared in a secure method. This algorithm will determine a frequency of the smallest and largest distances between the different neighbors.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Perez, Montiel, Zhao and Park. Perez teaches a machine learning model which is able to determine a user based on indirect metadata. Montiel teaches a multiple party system which is able to preserve the privacy of its users and is able to communicate with user without revealing personal information. Zhao teaches a method which uses privacy-preserving techniques and deep learning to generate collaborative user networks. Park teaches a method which is able to find nearest neighbors in an enclosed environment and a way to calculate different values between users, like distance. One of ordinary skill would have motivation to combine a machine learning model which is able to identify users based on their meta data with a privacy-persevering multi-party environment where users are able to communicate without sharing personal information with a system which uses deep learning for assistance with security to build a collaborative user environment while also preserving user privacy and a method used to find nearest neighbors along with different user data such as distance between users, "However, in order to utilize cloud computing for efficient data mining, the privacy problem should be solved first and foremost. In this paper, we focused on efficient PPkNN to realize classification as one of data mining tasks and proposed efficient PkNC to allow to be run in parallel. We demonstrated that the proposed PkNC is superior to existing PPkNNs in terms of security and efficiency. PkNC protects the privacy of dataset, query including kNN result, and further hides access patterns of dataset. For PkNC, we pro- posed SkLE/SkSE, SCP, and SCF as building blocks, and formally proved their security." (Park, Conclusion, pp. 64632).
Regarding claim 17, Perez, Montiel and Zhao fail to explicitly disclose, “wherein determining, by the first computing system, the first share of the set of predicted labels comprises, for each category in the set:” and “applying a function corresponding to the category to the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value to generate a first share of a predicted label corresponding to the category for the given user profile.”.
However, Park discloses, “wherein determining, by the first computing system, the first share of the set of predicted labels comprises, for each category in the set:” (Overview of SKLE and SKSE, pp. 64621; "However, SkLE and SkSE are carried out in parallel for each data or in common for all data (for example, computing the number of largest data and comparing the number to parameter k are carried out in common for all data). Furthermore, SkLE (resp., SkSE) find k largest (resp., smallest) elements in only one execution without changing the value of the highest (resp., lowest) element to the lowest (resp., highest) value." Each of the neighbors in the system are located and labeled accordingly. In this article they are labeled to find the largest and smallest kin distance from the model.)
“applying a function corresponding to the category to the frequency at which true labels that correspond to the category in the sets of true labels for user profiles in the k nearest neighbor user profiles are true labels of the first value to generate a first share of a predicted label corresponding to the category for the given user profile.” (Overview of SKLE and SKSE, pp. 64622; "As mentioned in Section II-A, DH has input ciphertexts and CSP has a decryption key. In SkLE (resp., SkSE), each element
e
i
has auxiliary data
K
i
,
P
i
, and
C
i
whose details are as follows.
K
i
indicates whether an element ei is one of k largest (resp., smallest) elements where
K
i
= 1 if an element
e
i
is one of k largest (resp., smallest) elements and
K
i
= 0 otherwise. Once an element
e
i
becomes one of k largest (resp., smallest) elements, it is irreversible. (i.e.,
K
i
= 0
→
1 but 1
↛
0).
P
i
indicates whether an element
e
i
is a predicted largest (resp., smallest) element which means the element predicted as a largest (resp., smallest) element where
P
i
= 1 if an element
e
i
is a predicted largest (resp., smallest) element and
P
i
= 0 otherwise." The model proposed in this article uses the data given to label the neighbors data based on distance. This model will take note of all nearest and farthest neighbors by distance.)
Claims 18 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Perez and Montiel in view of Xue et al., (Xue et al., “Secure and Privacy-Preserving Decision Tree Classification with Lower Complexity”, 2020, hereinafter “Xue”).
Regarding claim 18, Perez and Montiel fail to explicitly disclose, “wherein the client device computes the given user profile using a plurality of feature vectors that each include feature values related to events of a user of the client device and a decay rate for each feature vector.”
However, Xue discloses, “wherein the client device computes the given user profile using a plurality of feature vectors that each include feature values related to events of a user of the client device and a decay rate for each feature vector.” (System Model, pp. 18; "A client owns a private input which is represented as a feature vector that contains information of different attributes, such as weight, heart rate, blood pressure, etc. The client would like to leverage the model generated by the service provider to obtain the classification result of her input. Since the input contains sensitive information, the client would not offer the feature vector to the service provider in the plain text. In addition, due to the limited storage and computation resources of the client, the communication and computation overhead at the client should be small." In this model the user can use their own private deice to communicate with the network. The client device is able to communicate using feature vectors to keep their information private.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Perez, Montiel and Xue. Perez teaches a machine learning model which is able to determine a user based on indirect metadata. Montiel teaches a multiple party system which is able to preserve the privacy of its users and is able to communicate with user without revealing personal information. Xue teaches a method which uses different machine learning methods on local and global devices in a system to provide security and privacy-preserving techniques. One of ordinary skill would have motivation to combine a machine learning model which is able to identify users based on their meta data and a privacy-persevering multi-party environment where users are able to communicate without sharing personal information and a system which is able to use different machine learning models to provide security and protect against outside adversaries, "With the design of an efficient secure comparison protocol, the proposed decision tree classification scheme has achieved a lower computation and communication overhead for both the client and the service provider. Moreover, the formal security proof has demonstrated that the proposed scheme achieves the desired properties under the semi-honest model." (Xue, Conclusion, pp. 23-24).,
Regarding claim 19, Perez and Montiel fail to explicitly disclose, “wherein the client device computes the given user profile using a plurality of feature vectors that each include feature values related to events of a user of the client device, wherein computing the given user profile comprises:”, “classifying one or more of the plurality of feature vectors as sparse feature vectors; and” and “classifying one or more of the plurality of feature vectors as dense feature vectors, the method further comprising: generating the first share of the given user profile and respective second shares of the given user profile for the one or more second computing systems using the sparse feature vectors and dense feature vectors, wherein generating the first share and the respective second shares of the given user profile comprises splitting the sparse feature vector using a Function Secret Sharing (FSS) technique.”
However, Xue discloses, “wherein the client device computes the given user profile using a plurality of feature vectors that each include feature values related to events of a user of the client device, wherein computing the given user profile comprises:” (System Model, pp. 18; "A client owns a private input which is represented as a feature vector that contains information of different attributes, such as weight, heart rate, blood pressure, etc. The client would like to leverage the model generated by the service provider to obtain the classification result of her input. Since the input contains sensitive information, the client would not offer the feature vector to the service provider in the plain text. In addition, due to the limited storage and computation resources of the client, the communication and computation overhead at the client should be small." In this model the user can use their own private deice to communicate with the network. The client device is able to communicate using feature vectors to keep their information private.)
“classifying one or more of the plurality of feature vectors as sparse feature vectors; and” (System Model, pp. 18; "At a high level, our system works as follows. First, clients and the service provider generate their public-private keys and register with a TA, who will issue certificates for them. Then, a client encrypts his or her input vector and sends the ciphertext to the service provider. After receiving the input of the model, the service provider executes the decision tree classification on the ciphertexts and returns the protected result to the client, who can recover the classification result using his or her private key." This system discloses the use of input vectors into the system. These vectors contain user data and send cipher-text to the server.)
“classifying one or more of the plurality of feature vectors as dense feature vectors, the method further comprising: generating the first share of the given user profile and respective second shares of the given user profile for the one or more second computing systems using the sparse feature vectors and dense feature vectors, wherein generating the first share and the respective second shares of the given user profile comprises splitting the sparse feature vector using a Function Secret Sharing (FSS) technique.” (Secure Comparison Protocol, pp. 20; "In order for
P
1
and
P
2
jointly determine the comparison result, which means that
P
1
cannot learn the actual result without the information of
P
2
, we combine the secret sharing during the comparison. To be specific, after receiving the encryption of
v
1
,
P
2
randomly chooses a bit b
←
{0,1}." This article teaches a method which uses secure and privacy preserving techniques to comminate using vectors. This article teaches a secret sharing technique during comparison to not reveal secure data.)
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PAUL MICHAEL GALVIN-SIEBENALER whose telephone number is (571)272-1257. The examiner can normally be reached Monday - Friday 8AM to 5PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Viker Lamardo can be reached at (571) 270-5871. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/PAUL M GALVIN-SIEBENALER/Examiner, Art Unit 2147
/ERIC NILSSON/Primary Examiner, Art Unit 2151
Prosecution Insights