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 .
Response to Arguments
Applicant's arguments filed 11/28/2025 have been fully considered but they are not persuasive.
Regarding Objections to the Drawings in page 10 of remarks, the applicant states they submitted replacement sheet for FIG. 4. However no replacement sheet was received.
Regarding 101 arguments on page 11-12 of remarks, applicant argues “In the Office Action, claims 1-20 were rejected as allegedly directed to an abstract idea without significantly more. In particular, the Office Action alleges that elements of claims 1-20 may be interpreted as mental processes. ... In particular, the claims set forth an improvement in artificial intelligence systems. As recently reinforced by the decision directed by Director Squires in Ex parte Guillaume Desjardins et al., "an improvement to how the machine learning model itself operates" integrates any abstract idea into a patent eligible practical application. (Ex parte Guillaume Desjardins et al. page 9). For example, Para. [0109] of the specification describes: … Thus, the claims constitute an improvement to the computerized operation of a machine learning system. Such an improvement establishes a practical application of any abstract idea that is included in the claims. Accordingly, reconsideration of the rejections under 101 in light of the pending claims and arguments is requested.” The applicant argues amended limitations and how they constitute an improvement on the technology. The applicants main argument is directed towards the amended claims. The amended claims have not been examined and therefore the argument is moot and not persuasive. (Examiner notes that the applicant states how in the specification they provide improvement on the technology however the amended limitations do not explain why attestation is done. Adding additional limitations on how performing the attestation improves upon the technology could provide enough information to overcome 101.)
Regarding 103 rejection argumetns on page 13 of remarks, applicant argues “The claims set forth cause a training data set and training results received from a server to be cached in the memory; while the apparatus is disconnected from the server, train a machine learning model using the training data set received from the server to generate local training results; compare the local training results to training results received from the server; determine if attestation of the local training results passes based on the comparison of the local training results and the training results received from the server; and discard the machine learning model when attestation does not pass. The cited references do not teach or suggest such a process. In view of the amended claims, reconsideration is requested.” The applicants main argument is directed towards the amended claims. The amended claims have not been examined and therefore the argument is moot and not persuasive.
Drawings
The drawings are objected to under 37 CFR 1.83(a) because they fail to show:
Fig. 4. 440 points to the endpoint layer. The endpoint layer is referred to as 410 in the specification para 0035.
Fig. 4. Shows the edge gateway node as 442 the specification mentions 412 in para 0035.
as described in the specification. The applicant stated they submitted drawings in Applicant Request for Reconsideration-After dated 8-28-2025. However no drawings were received to amend Figure 4 as recited in page 2. Any structural detail that is essential for a proper understanding of the disclosed invention should be shown in the drawing. MPEP § 608.02(d). Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. The figure or figure number of an amended drawing should not be labeled as “amended.” If a drawing figure is to be canceled, the appropriate figure must be removed from the replacement sheet, and where necessary, the remaining figures must be renumbered and appropriate changes made to the brief description of the several views of the drawings for consistency. Additional replacement sheets may be necessary to show the renumbering of the remaining figures. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Claim Objections
Claims 8 is objected to because of the following informalities:
Claim 8 recites the “a golden training data” and “golden training results”. However both golden training data and training results are not used later in the claim. The limitation should be a training data set and training results as recited in claim 1.
Claim 9 recites the “attestation means are ”. The claim should recite “attestation means to determine”.
Appropriate correction is required.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that use the word “means,” and are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses means without reciting sufficient structure to perform the recited function. Such claim limitation(s) are part of Claim 8 and dependent claims 9, 10, and 13:
Claim 8 recites the limitations:
training means to cause a training data set and training results received from a server to be cached in the memory and while the apparatus is disconnected from the server.
attestation means to:
Claim 9 recites the limitation:
attestation means are
Claim 10 recites the limitation:
wherein the training means are to emulate a system
Claim 13 recites the limitation:
attestation means are
Claim 14 recites the limitation:
attestation means are
The specification in para 0048, 0050, 0054 provides the structure need to provide the means for the limitations mentioned above.
Para 0048: “FIG. 6 is a block diagram of an example implementation of an analyzer 600 to perform attestation of a model at a device (e.g., the client compute node 402) that is remote from a server (e.g., a server at the core data center 432).”
Para 0050: “The example model trainer 604 may be implemented by a processor executing instructions, an AI hardware accelerator, etc. ”
Para 0054: “While an example manner of implementing the example analyzer 600 is illustrated in FIG. 6, one or more of the elements, processes and/or devices illustrated in FIG. 6 may be combined, divided, re-arranged, omitted, eliminated and/or implemented in any other way. Further, the example interface 602, the example model trainer 604, the example parameter analyzer 606, the example attestation result generator 608, and the example data analyzer 610 and/or, more generally, the example analyzer 600 of FIG. 6 may be implemented by hardware, software, firmware and/or any combination of hardware, software and/or firmware. Thus, for example, any of the example interface 602, the example model trainer 604, the example parameter analyzer 606, the example attestation result generator 608, and the example data analyzer 610 and/or, more generally, the example analyzer 600 of FIG. 6 could be implemented by one or more analog or digital circuit(s), logic circuits, programmable processor(s), programmable controller(s), graphics processing unit(s) (GPU(s)), digital signal processor(s) (DSP(s)) application specific integrated circuit(s) (ASIC(s)), programmable logic device(s) (PLD(s)) and/or field programmable logic device(s) (FPLD(s)). When reading any of the apparatus or system claims of this patent to cover a purely software and/or firmware implementation, at least one of the example interface 602, the example model trainer 604, the example parameter analyzer 606, the example attestation result generator 608, and the example data analyzer 610 and/or, more generally, the example analyzer 600 of FIG. 6 is/are hereby expressly defined to include a non-transitory computer readable storage device or storage disk such as a memory, a digital versatile disk (DVD), a compact disk (CD), a Blu-ray disk, etc. including the software and/or firmware. Further still, the example analyzer 600 of FIG. 6 may include one or more elements, processes and/or devices in addition to, or instead of, those illustrated in FIG. 6, and/or may include more than one of any or all of the illustrated elements, processes and devices. As used herein, the phrase "in communication," including variations thereof, encompasses direct communication and/or indirect communication through one or more intermediary components, and does not require direct physical (e.g., wired) communication and/or constant communication, but rather additionally includes selective communication at periodic intervals, scheduled intervals, aperiodic intervals, and/or one time events. ”
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112 (d)
The following is a quotation of 35 U.S.C. 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 U.S.C. 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Claims 7, 14, and 21 are rejected under 35 U.S.C. 112(d) or pre-AIA 35 U.S.C. 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends. Claims 7 and analogous 14 and 21 depend upon claims 1 and analogous 8 and 15. Claim 1 and analogous 8 and 15 recite the limitation “discard the machine learning model when attestation does not pass.” The same limitation is recited in claims 7 and analogous 14 and 21 “discard the machine learning model if the attestation does not pass.” The limitation of claim 7 and analogous 14 and 21 does not further limit claim upon which they depend. Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
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 1-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to abstract idea without significantly more. The claim(s) recite(s) significantly more. The subject matter eligibility test for products and process is describe below for claim 1 in view of dependent claims.
Regarding claim 1:
Step 1: Is the claim to a process machine manufacture or composition of matter?
Yes – Claim 1 recites an apparatus, which is a machine that falls under the statutory categories.
Step 2A Prong 1: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes – The claim recites the following:
“compare the local training results to results received from the server; ” - The limitations of claim 1 recites a mental process of comparing the training results with the training results received from the server (see MPEP 2106.04(a)(2)III).
“[[and]] determine if attestation of the local training results passes based on the comparison of the local training results and the received from the server; ” - The limitation recites the mental process to determine if attestation passes (see MPEP 2106.04(a)(2)III).
Step 2 Prong 2: Does the claim recite additional elements that integrate the judicial exception into a particular application? No –
The claim includes the additional element(s):
“An apparatus for attesting a machine learning model, the apparatus comprising: memory; instructions; and at least one processor to execute machine readable instructions to at least: cause a training data set and training results received from a server to be cached in the memory;”
The additional elements fall under “apply it” as using a generic computer to train a machine learning model. See Mere Instructions to Apply an Exemption (see MPEP 2106.05(f)).
“while the apparatus is disconnected from the server, train a machine learning model using the training data set received from [[a]] the server to generate local training results; ”
The additional elements fall under “apply it” as using a generic computer to train a machine learning model while disconnected from the server. See Mere Instructions to Apply an Exemption (see MPEP 2106.05(f)).
and discard the machine learning model when attestation does not pass.
The additional elements fall under “apply it” as using a generic computer to discard the model when attestation does not pass. See Mere Instructions to Apply an Exemption (see MPEP 2106.05(f)).
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception?
No - The claim does not include additional elements that are sufficient to amount to a significantly more than the judicial exemption. As an order whole, the claim is directed towards attestation of a machine learning model. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of training and obtaining fall under using generic computer to apply an exemption and mere data gathering. The method does not improve on the function of a computer, transforms an article into another article, nor is it applied by a particular machine, making the claim not patent eligible.
Regarding claim 2:
Step 2A Prong 2, Step 2B: The additional element(s):
“The apparatus of claim 1, wherein the processor is to execute the instructions to determine if the attestation passes based on a difference between weights of the training and weights of the local training results.”
The additional elements fall under “apply it” as using a generic computer to execute instructions to determine if the attestation passes based on difference between weights . See Mere Instructions to Apply an Exemption (see MPEP 2106.05(f)).
Regarding claim 3:
Step 2A Prong 2, Step 2B: The additional element(s):
“The apparatus of claim 1 ,wherein the at least one processor is to execute the machine readable instructions to emulate a system state of the server during the training ”
The additional elements fall under “apply it” as using a generic computer to emulate a system. See Mere Instructions to Apply an Exemption (see MPEP 2106.05(f)).
Regarding claim 4:
Step 2A Prong 2, Step 2B: The additional element(s):
“The apparatus ”
The additional elements fall under “apply it” as using a generic computer to perform attestation in a multi-tier edge architecture . See Mere Instructions to Apply an Exemption (see MPEP 2106.05(f)).
Regarding claim 5:
Step 2A Prong 2, Step 2B: The additional element(s):
“The apparatus of claim1”
The additional elements fall under Insignificant Extra-Solution Activity. See MPEP 2106.5(g). The judicial exemptions do not integrate into a practical application nor provide an improvement. The process does not provide an inventive concept nor provides a practical application.
Regarding claim 6:
Step 2A Prong 2, Step 2B: The additional element(s):
“The apparatus of claim 1”
The additional elements fall under “apply it” as using a generic computer to perform attestation while disconnected from the network. See Mere Instructions to Apply an Exemption (see MPEP 2106.05(f))”
“download the machine learning model from the server while connected to a network; download the training samples from the server while connected to the network”
The additional elements fall under Insignificant Extra-Solution Activity as mere data gathering by downloading the model and training data samples. See MPEP 2106.5(g).
Regarding claim 7:
Step 2A Prong 2, Step 2B: The additional element(s):
“The apparatus of claim 1 wherein the at least one processor is to execute the machine readable instructions to discard the machine learning model if the attestation does not pass.”
The additional element falls under the “apply it” by using computers to execute instruction to discard a machine learning model if attestation does not pass (MPEP 2106.05(f)). The judicial exemptions do not integrate into a practical application nor provide an improvement. The process does not provide an inventive concept nor provides a practical application.
Claims 8-14 recite an apparatus and are analogous to the method of claims 1-7. Therefore, the rejections of claim 1-7 above applies to claims 8-14.
Claims 15-21 recite a computer readable medium product and are analogous to the method of claims 1-7. Therefore, the rejections of claim 1-7 above applies to claims 15-21.
Claims 22-25 recite a method and are analogous to the method of claims 1-5. Therefore, the rejections of claim 1-5 above applies to claims 22-25.
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.
Claim(s) 1, 4, 7, 8, 11, 14, 15, 18, 21, and 22 - 25 are rejected under 35 U.S.C. 103 as being unpatentable over Gu et al. (US20200082270A1) (“Gu”) in view of Nagaraju et al. (US20180032908A1) and further in view of H. Chen, C. Fu, B. D. Rouhani, J. Zhao and F. Koushanfar, "DeepAttest: An End-to-End Attestation Framework for Deep Neural Networks," 2019 ACM/IEEE 46th Annual International Symposium on Computer Architecture (ISCA), Phoenix, AZ, USA, 2019, pp. 487-498 (“Chen”).
Regarding claim 1 and analogous claims 8, 15 and 22 Gu teaches an apparatus for attesting a machine learning model, the apparatus comprising (Gu Para 0048, FIG. 1 is an example diagram illustrating a training stage workflow and interaction of operational components of a verifiable deep learning training cloud service in accordance with one illustrative embodiment. In the depiction in FIG. 1, the client side operations are performed on collaborative participant computing devices or data processing systems 110, 120, while the server side operations are performed on one or more server computing devices or data processing systems 130 implementing the verifiable deep learning training service infrastructure of the illustrative embodiments.
Para 0030 line 1-13, In accordance with the illustrative embodiments, the training operation for training a deep learning pipeline, or machine learning (ML) model, is also split into FrontNet subnet model training and BackNet subnet model training. The FrontNet subnet model training is executed in an isolated trusted execution environment (TEE) with memory access control and encryption enforcement. The users only need to provision encrypted training data to the training service providers. Thus, no one other than the end users can inspect the content of the training data outside of the TEE. The processes running inside of the TEE authenticate the source of the training data, decrypt the encrypted training data, and verify its validity.
Para 0031, Thus, the training processes and unencrypted original training data provided by the various training data sources for training the FrontNet subnet are kept within the perimeter of a specific TEE and are invisible to the external world. Furthermore, the TEE can attest to remote parties (i.e., the end users of the cloud training services) that the FrontNet subnet model is running in a secure environment hosted by a trusted hardware platform [An apparatus for attesting].
Para 0102, For each received encrypted training dataset, at the TEE, the received encrypted dataset from the contributor is authenticated with the provisioned security key for that contributor and the integrity of the encrypted dataset is checked (step 530). If any check fails, the encrypted training dataset is discarded (step 535). If all checks are passed, then the received encrypted training dataset is decrypted and processed via a FrontNet subnet model of the deep learning model executing within the TEE (step 540). The FrontNet subnet model passes intermediate representations (IRs) to the BackNet subnet model executing outside the TEE which processes the IRs to generate an output result (step 545). Backpropagation and weight updating is performed to train the deep learning model (step 550) and a determination is made as to whether the output result is sufficient to warrant discontinuing the training of the deep learning model (step 555). If not, then the operation is repeated with the next iteration of training data. If the training has completed, then the fingerprints for the dataset instances are generated based on the trained deep learning model (step 560) and the trained deep learning model is released to the training dataset contributors (step 565). The operation then terminates.
(Examiner Note: That the specification in para 0047 calls for verification of the machine learning model. Specification Para 0047, “Example attestation approaches disclosed herein utilize a golden set of data to validate a machine learning model (e.g., CNN model) being trained and operated at a remote system”)):
memory;
instructions;
and at least one processor to execute machine readable instructions to at least (Gu Para 0098, As mentioned above, in some illustrative embodiments the mechanisms of the illustrative embodiments may be implemented as application specific hardware, firmware, or the like, application software [instructions;] stored in a storage device [machine readable instructions], such as HDD 426 and loaded into memory, such as main memory 408 [memory;], for executed by one or more hardware processors [one processor], such as processing unit 406, or the like. As such, the computing device shown in FIG. 4 becomes specifically configured to implement the mechanisms of the illustrative embodiments and specifically configured to perform the operations and generate the outputs described herein with regard to the deep learning cloud service implementing the privacy enhancing deep learning cloud service framework and one or more processing pipelines.):
cause a training data set and training results received from a server to be cached in the memory (Gu para 0073 line 1-18, As shown in FIG. 3A, one or more of the computing devices, e.g., server 304A, may be specifically configured to implement a deep learning cloud service 300 which further implements a verifiable deep learning training service framework 320, in accordance with one illustrative embodiment. The configuring of the computing device may comprise the providing of application specific hardware, firmware, or the like to facilitate the performance of the operations and generation of the outputs described herein with regard to the illustrative embodiments. The configuring of the computing device may also, or alternatively, comprise the providing of software applications stored in one or more storage devices and loaded into memory of a computing device, such as server 304A, for causing one or more hardware processors of the computing device to execute the software applications that configure the processors to perform the operations and generate the outputs described herein with regard to the illustrative embodiments.
Para 0084, The automated fingerprint generation module 334 performs operations for generating fingerprints and evidence data for the training dataset instances used to train the deep learning model 350 which is then stored in the evidence storage 332 for later processing of queries by the query module 334. The operations of the security module 322, data augmentation module 324, and training logic module 328 are described in FIG. 1 above. The operations of the automated fingerprint generation module 330, evidence storage 332, and query module 334 are described in FIG. 2 above. It should be appreciated that the client computing devices 310, 312 from which training data is received, in some illustrative embodiments, may themselves be servers or other types of computing devices used at an organizational level of operation, rather than individual client computers associated with individual persons [a training data].
Para 0088, The encrypted input data is decrypted, such as by the security module 322, to generate the original input data. The input data is input to the FrontNet subnet model 352 of the trained deep learning model 306 which generates intermediate representations (IR) that are output to the BackNet subnet model 354. The BackNet subnet model 354 then processes the IR output from the FrontNet subnet model 352 to generate a runtime result, e.g., a classification output or the like, that is provided back to the deep learning cloud service 300 and/or processing pipeline 305 for use in performing a deep learning operation based on the input data. Results of the deep learning operation [training results]. (i.e. the training data and training results are cached by the server));
compare the local training results to results received from the server (Gu para 0102, For each received encrypted training dataset, at the TEE, the received encrypted dataset from the contributor is authenticated with the provisioned security key for that contributor and the integrity of the encrypted dataset is checked (step 530). If any check fails, the encrypted training dataset is discarded (step 535). If all checks are passed, then the received encrypted training dataset is decrypted and processed via a FrontNet subnet model of the deep learning model executing within the TEE (step 540). The FrontNet subnet model passes intermediate representations (IRs) to the BackNet subnet model executing outside the TEE which processes the IRs to generate an output result (step 545). Backpropagation and weight updating is performed to train the deep learning model (step 550) and a determination is made as to whether the output result is sufficient to warrant discontinuing the training of the deep learning model (step 555). If not, then the operation is repeated with the next iteration of training data. If the training has completed, then the fingerprints for the dataset instances are generated based on the trained deep learning model (step 560) and the trained deep learning model is released to the training dataset contributors (step 565));
[[and]] determine if attestation of the local training results passes based on the comparison of the local training results and the received from the server ((Gu para 0062, To address the accountability issue, the illustrative embodiments provide a fingerprinting mechanism 220, 242 to discover the poisoned and/or mislabeled training datasets 210 that lead to an erroneous output during runtime operation, e.g., a runtime misclassification of input data X by the trained deep learning model 240. Instead of retaining the original training data for runtime inspection, the fingerprinting mechanism 220, 242 of the illustrative embodiments records evidence 225 for each training data instances 210, which may then be used to trace the data source and training dataset leading to the erroneous training of the deep learning model 150 [determine if attestation of the local training results passes].
para 0084, The automated fingerprint generation module 330 performs operations for generating fingerprints and evidence data for the training dataset instances used to train the deep learning model 350 which is then stored in the evidence storage 332 for later processing of queries by the query module 334. The operations of the security module 322, data augmentation module 324, and training logic module 328 are described in FIG. 1 above. The operations of the automated fingerprint generation module 330, evidence storage 332, and query module 334 are described in FIG. 2 above. It should be appreciated that the client computing devices 310, 312 from which training data is received, in some illustrative embodiments, may themselves be servers or other types of computing devices used at an organizational level of operation, rather than individual client computers associated with individual persons.
Para 0089. During this runtime operation, should an end user determine that the trained deep learning model 306 is generating erroneous outputs, e.g., erroneous classifications, the end user may submit the erroneous output Y and the fingerprint F generated for the new data processed by the trained pipeline 350 in a query to the deep learning cloud service 300. In response to receiving such a query, the query module 334 may search for similar fingerprints F with the same corresponding output Y in the evidence storage 332 [comparison of the local training results and the received from the server].);
Gu does not explicitly teach while the apparatus is disconnected from the server, train a machine learning model using the training data set received from [[a]] the server to generate local training results;
and discard the machine learning model when attestation does not pass.
However Nagaraju teaches while the apparatus is disconnected from the server, train a machine learning model using the training data set received from [[a]] the server to generate local training results (Nagaraju para 0023, During operation, at least some of the edge devices can be communicatively connected from the server computer system intermittently. Such a connection can be via a wired or wireless link or via a combination of wired and wireless links, and can be direct or indirect. Edge data and model data can be exchanged between an edge device and the server computer system while the edge device is connected to the server computer system. The edge device can continue to operate using the local model even when the edge device is disconnected from the server computer system. In some embodiments, the edge device can use a local machine-learning process to train its local model when the edge device is disconnected from the server computer system. In some embodiments, any of the edge devices or server computer systems can use schemas to extract training data from unprocessed data or minimally processed data (“raw data”) or data derived from the raw data [while the apparatus is disconnected from the server, train a machine learning model].
Para 0121, In step 1008, the raw data and/or data items are processed with a local model to produce output data. In step 1010, the edge device may execute a local action as a function of the output data. For example, the edge device can change a temperature setting (local action) of a thermostat as a result of processing temperature data (raw data) with the local model.
Para 0124, In step 1016, the edge device receives model data from the server computer system. The model data is derived from the training data collected from the edge devices (“global training data”) at the server computer system. In particular, the global training data is input to one or more machine-learning processes (e.g., algorithms) to generate a global model. As such, the global model is trained by harmonizing local data from the edge devices. The model data can include the global model or data related to the global model, such as parameters indicative of changes necessary to synchronize a local model with the global model.
Para 0125 In step 1018, the local model of the edge device is updated with the model data. Hence, the local model can be synchronized with the global model such that local data is subsequently processed with the updated local model in light of a global context. In some embodiments, the local model is replaced with the global model or updated with parameter values that modify the local model accordingly [using the training data set received from [[a]] the server to generate local training results]);
Gu and Nagaraju are considered to be analogous to the claim invention because they are in the same field of distributed machine learning. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to have modified Gu to incorporate the teachings of Nagaraju by training a machine learning model while the server is disconnected from the service. Doing so to ensure the edge devices can operate autonomously or semi-autonomously to carry out designated task (Nagaraju para 0027 At least some of the edge devices 12 may be only intermittently connected from the networks 16. As shown in FIG. 1, for example, edge device 12-1 is disconnected from the network 16 while edge device 12-2 is connected to the network 16. The edge devices 12 can generate edge data and provide the edge data to the server computer system 14 over the networks 16. In some embodiments, the edge devices 12 can operate autonomously or semi-autonomously to carry out designated tasks. In some embodiments, the edge devices 12 can operate under the control of a user to carry out tasks.
Para 0028 line 1-3, Each of the edge devices 12 can generate edge data locally based on inputs received by the respective edge devices 12, according to their designed functionality.).
Chen teaches and discard the machine learning model when attestation does not pass (Chen Page 4, 3.1 DNN Attestation Metrics. We introduce a comprehensive set of criteria to evaluate the performance of a DNN attestation technique. Table 1 details the criteria for an effective DNN attestation methodology. Fidelity requires that the functionality (e.g., accuracy) of the pre-trained model shall not be degraded after the off-line DNN marking. Reliability and integrity means that the attestation approach shall prevent unauthorized DNNs from executing (low false alarm rate of FP detection) [the machine learning model if the attestation does not pass] and allow normal inference of legitimate DNNs (high detection rate of the embedded FP), respectively. A reliable attestation method is also desired to satisfy the security requirement such that the attestation decision is trustworthy. Efficiency requires that the overhead (i.e., latency, power consumption) incurred by attestation shall be negligible. Scalability and generalizability ensure that the attestation method can be applied to DNNs of various size and diverse TEE-supported hardware devices, respectively. DeepAttest satisfies all the requirements listed in Table 1 as shown in Section 6. (i.e. machine learning models that do not pass attestation are discarded from executing)).
Gu and Chen are considered to be analogous to the claim invention because they are in the same field of machine learning. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to have modified Gu to incorporate the teachings of Chen of preventing the execution of models that fail attestation. Doing so ensure that authorized DNN programs are able to interface on the target device (Chen Abstract line 16-23, The existence of the pre-defined FP is used as the attestation criterion to determine whether the queried DNN is authenticated. Our attestation framework ensures that only authorized DNN programs yield the matching FP and are allowed for inference on the target device. DeepAttest provisions the device provider with a practical solution to limit the application usage of her manufactured hardware and prevents unauthorized or tampered DNNs from execution.).
Regarding claim 4 and analogous claims 11, 18 and 25, Gu in view of Nagaraju and Chen teach the apparatus of claim 1 analogous claims 8, 15 and 22.
Gu teaches wherein the attestation is performed in a multi-tier edge architecture (Gu Para 0075, As shown in FIG. 3A, one or more of the servers 304A-304C are configured to implement the deep learning cloud service 300 and verifiable deep learning training service framework 320 (hereafter referred to as the "framework" 320). While FIG. 3A shows elements 300 and 320 being associated with a single server, i.e. server 304A, it should be appreciated that a plurality of servers, e.g., 304A-304C, may together constitute a cloud computing system and be configured to provide the deep learning cloud service 300 implementing the framework 320 such that the mechanisms of the deep learning cloud service 300, including the framework 320 or portions thereof, and the processing pipeline(s) 305 or portions thereof, may be distributed across multiple server computing devices 304A-304C. In some illustrative embodiments, multiple instances of the deep learning cloud service 300, pipeline(s) 305, and frame work 320 may be provided on multiple different servers 304A-304C of the cloud computing system [a multi-tier edge architecture]. The deep learning cloud service 300 may provide any deep learning or AI based functionality of a deep learning system, an overview of which, and examples of which, are provided hereafter. (See Fig. 3A-B)).
Regarding claim 7 and analogous claims 14 and 21, Gu in view of Nagaraju and Chen teach the apparatus of claim 1 analogous claims 8, 15 and 22.
Gu, Nagaraju and Chen are combined in the same rational as set forth above with respect to claim 1 and analogous claims 8, 15 and 22.
Chen teaches, wherein the at least one processor is to execute the machine readable instructions to discard the machine learning model if the attestation does not pass (Chen Page 490, 3.1 DNN Attestation Metrics. We introduce a comprehensive set of criteria to evaluate the performance of a DNN attestation technique. Table 1 details the criteria for an effective DNN attestation methodology. Fidelity requires that the functionality (e.g., accuracy) of the pre-trained model shall not be degraded after the off-line DNN marking. Reliability and integrity means that the attestation approach shall prevent unauthorized DNNs from executing (low false alarm rate of FP detection) [the machine learning model if the attestation does not pass] and allow normal inference of legitimate DNNs (high detection rate of the embedded FP), respectively. A reliable attestation method is also desired to satisfy the security requirement such that the attestation decision is trustworthy. Efficiency requires that the overhead (i.e., latency, power consumption) incurred by attestation shall be negligible. Scalability and generalizability ensure that the attestation method can be applied to DNNs of various size and diverse TEE-supported hardware devices, respectively. DeepAttest satisfies all the requirements listed in Table 1 as shown in Section 6. (i.e. machine learning models that do not pass attestation are prevented from executing and effectively discarding)
Page 493, Early Termination. To further reduce the attestation overhead, we avoid unnecessary computation and communication using early termination. More specifically, the online attestation terminates and yields the abortion command once a mismatch between the extracted FP segment and the pre-specified device-specific FP is detected as shown in Figure 5.).
Regarding claim 22-25, Gu in view of Nagaraju and Chen teach the apparatus of claim 1 analogous claims 8, 15 and 22 as mentioned above. However claim 22 includes a condition that even without prior art the claims are being rejected as MPEP 2111.04 recites “The broadest reasonable interpretation of a method (or process) claim having contingent limitations requires only those steps that must be performed and does not include steps that are not required to be performed because the condition(s) precedent are not met.” Claim 22 recites a method that does not explain what happens if the attestation passes. The current condition “when attestation does not pass” would allow one in the art to be able to use the conditions in case they are not triggered. Thus claim 22-25 would be rejected.
Claim(s) 2, 9, 16 and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Gu in view of Nagaraju and Chen and further in view of Ghahramani et al. (US20180114113A1) (“Ghahramani”).
Regarding claim 2 and analogous claims 9, 16, and 23, Gu in view of Nagaraju and Chen teach the apparatus of claim 1 analogous claims 8, 15 and 22.
Gu, Nagaraju and Chen are combined in the same rational as set forth above with respect to claim 1 and analogous claims 8, 15 and 22.
Gu do not explicitly teach wherein the processor is to execute the instructions to determine if the attestation passes based on a difference between weights of the training and weights of the training results
However Ghahramani teaches wherein the processor is to execute the instructions to determine if the attestation passes based on a difference between weights of the training and weights of the training results (Ghahramani Para 0022, A direct mapping for a network is learned in conjunction with an indirect network that designates expected weights for the direct mapping . The network generating the “ direct mapping ” may also be termed a "direct network ” or a “ direct model . ” The indirect network learns an expected weight distribution of the weights of the direct network , which may be represented as a set of "expected ” weights for the direct mapping . The indirect network may also be termed an “ indirect model . ” The direct model may include a portion of a larger modeled network , such as a multi - node, multi - layered neural network , wherein each direct network models transitions for one or more nodes in the network.
Para 0025, In one approach , the indirect model is used to regularize the weights applied to the mapping in the direct network . In this approach , when training the model, the error term for the direct weights is regularized by the expected weights given by the indirect network. In this way , the indirect network provides an ' anchor ' or set point from which the direct network weights may vary when it more accurately reflects the data [based on a difference between weights of the training and weights of the training results]).
Gu and Ghahramani are considered to be analogous to the claim invention because they are in the same field of machine learning. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to have modified Gu to incorporate the teachings of Ghahramani of determining the difference between weights. Doing so to capture properties relevant to the data which may not be known (Ghahramani Para 0020, Finally , other conventional regularizers are manually configured and don't adapt to the data. An intelligent, adaptive regularization system must be able to capture properties relevant to the data which may not be known a priori).
Claim(s) 3, 10, 17, and 24 are rejected under 35 U.S.C. 103 as being unpatentable over Gu in view of Nagaraju and Chen and further in view of Bendre et al. (US10380504B2) (“Bendre”).
Regarding claim 3, and analogous claims 10, 17, and 24, Gu in view of Nagaraju and Chen teach the apparatus of claim 1 analogous claims 8, 15 and 22.
Gu, Nagaraju and Chen are combined in the same rational as set forth above with respect to claim 1 and analogous claims 8, 15 and 22.
Gu do not explicitly teach wherein the at least one processor is to execute the machine readable instructions to emulate a system state of the server during the training by the model trainer.
However Bendre teaches wherein the at least one processor is to execute the machine readable instructions to emulate a system state of the server during the training by the model trainer (Bendre Col 6 line 61-67, FIG. 1 is a simplified block diagram exemplifying a computing device 100, illustrating some of the components that could be included in a computing device arranged to operate in accordance with the embodiments herein. Computing device 100 could be a client device (e.g., a device actively operated by a user), a server device (e.g., a device that provides computational services to client devices),
Col 9 line 45-59, Virtual machines 308 may be embodied by one or more of computing device 100 or server cluster 200. In general, a virtual machine is an emulation of a computing system, and mimics the functionality ( e.g., processor, memory, and communication resources) of a physical computer [emulate a system state of the server]. One physical computing system, such as server cluster 200, may support up to thousands of individual virtual machines. In some embodiments, virtual machines 308 may be managed by a centralized server device or application that facilitates allocation of physical computing resources to individual virtual machines, as well as performance and error reporting. Enterprises often employ virtual machines in order to allocate computing resources in an efficient, as needed fashion. Providers of virtualized computing system include.
Col 18 line 12-19, Additionally, processor 612 configured to send an ML training request to the scheduler device 604, which, as further discussed herein, effectively triggers assignment of an ML trainer process to generate an ML model based on the solution definition. Furthermore, the processor 612 may be configured to receive a generated ML model from the trainer device 606 and to store that ML model within the customer instance 610 [during the training by the model trainer]
Col 35 line 51-54, The program code can include one or more instructions executable by a processor for implementing specific logical operations or actions in the method or technique.).
Gu and Bendre are considered to be analogous to the claim invention because they are in the same field of machine learning. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to have modified Gu to incorporate the teachings of Bendre of emulating the server. Doing so facilitate allocation of physical computing resources and error reporting (Bendre Col 9 line 51-57, In some embodiments, virtual machines 308 may be managed by a centralized server device or application that facilitates allocation of physical computing resources to individual virtual machines, as well as performance and error reporting. Enterprises often employ virtual machines in order to allocate computing resources in an efficient, as needed fashion.).
Claim(s) 5, 12, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Gu in view of Nagaraju and Chen and further in view of BOS (US2020005766A1) (“BOS”).
Regarding claim 5 and analogous claims 12 and 19, Gu in view of Nagaraju and Chen teach the apparatus of claim 1 analogous claims 8, 15 and 22.
Gu, Nagaraju and Chen are combined in the same rational as set forth above with respect to claim 1 and analogous claims 8, 15 and 22.
Gu do not explicitly teach wherein the apparatus is an edge compute node.
However BOS teach wherein the apparatus is an edge compute node (BOS Para 0018, FIG. 2 illustrates data processing system 20 for use in either IoT edge node 14 [an edge compute node] or IoT device 12 in accordance with an embodiment. Data processing system 20 may be implemented on one or more integrated circuits and may be used to implement either or both of machine learning unit 16 and secure element 18. Data processing system 20 includes bus 22.
Para 0023 line 1-12, FIG. 3 illustrates method 40 for remotely detecting tampering of a machine learning model in accordance with an embodiment. Machine learning models may be valuable assets. The ability to make an almost identical copy of a machine learning model by simple remote queries to the model is a growing problem for the owners of models. Also, tampering with the internal functionality of the machine learning models can cause incorrect output values with potentially harmful effects. Method 40 provides a method to detect if an attacker has tampered with a machine learning model. Method 40 may be implemented, for example, in the data processing system 20 of FIG. 2.).
Gu and BOS are considered to be analogous to the claim invention because they are in the same field of machine learning. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to have modified Gu to incorporate the teachings of BOS and determine if the model has been tampered with. Doing so to verify the model remotely without direct access to the model (BOS Para 0010, By training the model with an invalid input value, the integrity of a machine learning model can be verified remotely, without requiring direct local access to the model. The use of an invalid input value makes it more unlikely that an attacker will be able to guess or find the correct invalid input value that was used in the training phase.)
Claim(s) 6, 13 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Gu in view of Nagaraju and Chen and further in view of Ralhan et al. (US20190354809A1) (“Ralhan”) and Nagaraj et al. (US20130238534A1) (“Nagaraj”).
Regarding claim 6 and analogous claims 13 and 20, Gu in view of Nagaraju and Chen teach the apparatus of claim 1 analogous claims 8, 15 and 22.
Gu, Nagaraju and Chen are combined in the same rational as set forth above with respect to claim 1 and analogous claims 8, 15 and 22.
Gu does not explicitly teach wherein the at least one processor is to execute the machine readable instructions to: download the machine learning model from the server while connected to a network; download the training samples from the server while connected to the network; and perform attestation while disconnected from the network.
However Ralhan teaches wherein the at least one processor is to execute the machine readable instructions to: download the machine learning model from the server while connected to a network (Ralhan Para 0105, Some systems may use Hadoop®, an open-source framework for storing and analyzing big data in a distributed computing environment. Apache™ Hadoop® is an open-source software framework for distributed computing. For example, some grid systems may be implemented as a multi-node Hadoop® cluster, as understood by a person of skill in the art. Some systems may use cloud computing, which can enable ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction [from the server while connected to a network].
Para 0120, Model management framework 1102 may interface with data sources 1140, data process instances 1130, modeling instances 1132, model validation instances 1134, deployment instances 1136, and/or data storage 1170, and/or interfaces 1150, such as management micro services 1154. In some embodiments, deployment instance 1136 may interact with prediction instances 1138 and/or interfaces 1150, such as cognitive micro services.
Para 0149, Logic flow 2000 may list the model version from the model version topic at 2002, select the model for validation and download the model at block 2004 [download the machine learning model], validate the model at block 2006, update the model meta for validation status/results at block 2008, and list the model version and meta in the model version topic at block 2010.));
download the training samples from the server while connected to the network (Ralhan Para, 0105, Some systems may use Hadoop®, an open-source framework for storing and analyzing big data in a distributed computing environment. Apache™ Hadoop® is an open-source software framework for distributed computing. For example, some grid systems may be implemented as a multi-node Hadoop® cluster, as understood by a person of skill in the art. Some systems may use cloud computing, which can enable ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction [from the server while connected to a network].
Model management framework 1102 may interface with data sources 1140, data process instances 1130, modeling instances 1132, model validation instances 1134, deployment instances 1136, and/or data storage 1170, and/or interfaces 1150, such as management micro services 1154. In some embodiments, deployment instance 1136 may interact with prediction instances 1138 and/or interfaces 1150, such as cognitive micro services.
Para 0149, Logic flow 1900 may download vectorized rank training data at block 1902 [download the training samples], run a training process at block 1904, generate new rank model file at block 1906, upload new rank model file with meat at block 1908, and list the model version and meta in the model version topic at block 1910.);
Gu and Ralhan are considered to be analogous to the claim invention because they are in the same field of machine learning services. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to have modified Gu to incorporate the teachings of Ralhan and download a model and training data in a distributed system for validation. Doing so to provide more precise evaluation for specific user or business case (Ralhan 0111, In various embodiments, once challenger models are selected during model building 804, a challenger model process 822 may select challenger models to be provided to a continuous integration (CI)/continuous delivery (CD) process or pipeline 824 for deployment in a quality assurance 806 process. A model assessment process 808 may operate to assess the challenger models 810, for example, compared with the performance of champion models 812, using various metrics, including, without limitation, performance metrics, industrial metrics, user metrics, and/or the like (see, for example, FIG. 9). In various embodiments, user metrics 948 may be associated with performance for a particular user, user environment, files/documents, and/or the like. For example, user metrics 948 may be associated with whether a model working better for a user's particular documents, environment, and/or the like. Accordingly, models evaluated using user metrics 948 may provide more precise evaluations for a specific user, class of information (or documents), business case (or business control), and/or the like.).
Nagaraj teaches and perform attestation while disconnected from the network (Nagaraj Para 0012 9-16, The method may then include offline training, evaluating, and validating of a model configured for learning in areas of the service access quality of experience (QoE) issues, based on the output set, and utilizing the trained, evaluated, and validated model [perform attestation] to execute a prediction function to provide the offline root cause recommendations for the service access quality of experience (QoE) issues.
Para 32, FIG. 2 illustrates the functionality of the offline module 100 and the online prediction module 110, according to one embodiment. According to this embodiment, online prediction module 110 collects network data including FCAPS data, prepares predictor variables, performs aggregation to generate samples at a chosen interval, and prepares and provides the input data to offline module 100. From session logs, offline module 100 extracts the de-registration entries. Offline module 100 may then perform aggregation to generate samples at set intervals, perform high level QoE service access categorization, prepare output data, and prepare a training and validation evaluation set for each category. Offline module 100 may then execute a training function, execute a validation function, execute an evaluation function, and update the model [while disconnected from the network]).
Gu and Nagaraj are considered to be analogous to the claim invention because they are in the same field of machine learning services. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filling date of the claimed invention to have modified Gu to incorporate the teachings of Nagaraj perform validation of the model offline. Doing so to provide offline root cause recommendations for the service (Nagaraj Para 0013, The apparatus is also caused to offline train, evaluate, and validate a model configured for learning in areas of the service access quality of experience (QoE) issues, based on the output set, and to utilize the trained, evaluated, and validated model to execute a prediction function to provide the offline root cause recommendations for the service access quality of experience (QoE) issues.).
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 ALFREDO CAMPOS whose telephone number is (571)272-4504. The examiner can normally be reached 7:00 - 4:00 pm M - F.
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, Michael J. Huntley can be reached at (303) 297-4307. 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.
/ALFREDO CAMPOS/Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129
Prosecution Insights