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 .
DETAILED ACTION
The communication received on 2/6/26 has been entered.
Response to Arguments/Amendments
Applicant’s acknowledgement that the double patenting rejection would be addressed in the future is acknowledged.
Applicant's arguments have been carefully considered but they were not found persuasive.
Argument I
Rane’s data structure that is encrypted is not a machine learning data structure generated by training a machine learning model and it does not contain disclosure of training a machine learning model, generating a machine learning data structure from such training, or encrypting that machine learning data structure for transmission to a server.
Rane, the client's search data is encrypted for privacy; the operational model on the server side is not encrypted
Response I
In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
The argued limitations were addressed with additional prior art and motivation to combine.
Argument II
Rane does not teach the model evaluating data structure being machine learning model.
Response II
Again, in response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
The examiner refers applicant to Gilad’s reference and the motivation to combine.
Argument III
Gilad does not teach or suggest an "encrypted machine learning data structure formed by using a homomorphic encryption scheme to encrypt a machine learning data structure that has been generated by training a machine learning model”.
Response III
In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
The argued limitations were addressed with Rane in view of Rane, Gilad and Mitchell.
Argument IV
Mitchell does not disclose homomorphic encryption, encrypted machine learning data structures, trusted or untrusted environments, or any mechanism for protecting the intellectual property of trained machine learning models through encryption.
Response IV
In response to applicant's arguments against the references individually, one cannot show non-obviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
The argued limitations were addressed with combination of the prior art references.
Argument V
The Examiner further asserts that "a skilled in the art would readily appreciate that there are only two obvious environment choices within which to operate on data: trusted or non-trusted environment, selection of any of them offering the predictable benefit of customization "trusted environment" in context. At paragraph [0070], the specification states that "the client(s) 510 can operate in a secure or trusted environment"
The "trusted environment" in the claims is not merely a label for where computation occurs. The entire purpose of the homomorphic encryption in the claims is to protect the model's intellectual property when the model is deployed to an untrusted environment for evaluation.
Response V
Applicant provided guidance in the specification that “the client … can operate in a secure or trusted environment” following the fact that in untrusted environment (operating on unencrypted data (“an unencrypted machine analytic could be evaluated to learn information about computing being performed …” ) only emphasize examiner’s point of view.
In fact, giving the fact that Michell’s as modified invention utilizes homomorphic encryption, which a skilled in the art would readily appreciate as the encryption providing additional security to processing of (already) encrypted data, the Michell’s environment could reasonably be equate to “trusted environment”.
However, the main point that the examiner raised in the previous communication was the fact that the term “trusted environment” as used in the claim language, would not affect the invention as claimed and that it would have been one of the two obvious variants. Moreover, as used currently in the claim language, could be interpreted as just a label and applicant’s arguments as the intended use.
This being said, the examiner noted that in light of the claim language not putting additional functional language further limiting the invention by the use of this specific variant, using either one of particular environment (trusted/untrusted) would have been just an obvious variant offering the predictable benefit of customization (the use of the homomorphic encryption already secures the environment) and extra security.
Argument VI
Applicant traverses the Official Notice citing in the previous Office Action and request affidavit or citation of reference.
Response VI
Regardfully, applicant failed to adequately support the allegation.
While choosing selectively MPEP citations (MPEP 2144.03 (B)) applicant overlooked applicant’s responsibility in regard to the Official Notice that follows (see MPEP 2144.03 (C)) applicant’s cited MPEP fragments.
Applicant is reminded that to adequately traverse such a finding, an applicant must specifically point out the supposed errors in the examiner’s action, which would include stating why the noticed fact is not considered to be common knowledge or well-known in the art. See 37 CFR 1.111(b). See also Chevenard, 139 F.2d at 713, 60 USPQ at 241 (“[I]n the absence of any demand by appellant for the examiner to produce authority for his statement, we will not consider this contention.”).
As clearly cited in MPEP 2144.03 (C), a general allegation that the claims define a patentable invention without any reference to the examiner’s assertion of official notice would be inadequate and in case of applicant’s failure to traverse the examiner’s assertion of official notice or offer inadequate traversal, the common knowledge or well-known in the art statement is taken to be admitted prior art.
Argument VII
The Examiner asserts that it would have been obvious to combine Rane and Gilad for "the benefit of scalable and efficient computation."
This rationale does not explain how or why a person of ordinary skill would have been motivated to modify the combination to encrypt a machine learning model's data structures, as opposed to encrypting user data (which is what both Rane and Gilad actually teach).
Response VII
Applicant’s argument is not well understood. An additional prior reference was not to suggests adding one encryption to another. The purpose of introducing Gilad’s teaching was to illustrate that the concept of encrypted data could be extended to various type data, such as machine leaning model, the solution enabling scalability, especially in light of the continues/increasing implementation ML in computing.
Argument VII
Arguments towards the dependent claims are substantially similar to the previously discussed arguments
Response VII
Applicant is referred to the previously offered response, above.
Claims 1-20 are pending.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Double Patenting
Claim(s) 1-20 remain rejected on the ground of non-statutory double patenting for the reasons of record. (See the details in the previous Office Action.)
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
Claim Rejections - 35 USC § 103
Claim(s) 1-2, 4-9, 11-12 and 14-19, remain rejected under 35 U.S.C. 103 unpatentable over Rane (USPUB 20160182222) in view of Gilad (USPUB 20160350648) and further in view Mitchell (USPUB 20190266282).
As per claims 1-2, 4, 7-9, 11-12, 14 and 17-19, in Fig. 3 and the associated text, Rane teaches receiving, from a client, by at least one server in an environment, an encrypted data structure formed by using a fully homomorphic encryption scheme to encrypt a data structure (client 11 sends to the server 14 query vector data results encrypted with the fully homomorphic encryption, para 47, Fig. 3 step 55-56); extracting, by the at least one server, a previously unseen instance of data, evaluating, by the at least one server, the encrypted data structure over the previously unseen instance of data using the machine learning model containing the encrypted data structure to generate at least one encrypted result about the previously unseen instance of data (Fig. 3 step 57); and sending, from the at least one server, the at least one encrypted result to the client (server applies a homomorphic encryption of a result and provides the encrypted results to the client, para 47 and 51-52 and claim 11, Fig. 3 steps 57-59), the at least one encrypted result configured to be decrypted at the client using the homomorphic encryption scheme (Fig. 3 step 61).
Rane does not teach the model evaluating data structure being machine learning model. However, in the related art, Gilad suggests utilizing such model (the neural network computation component perform computation on the encrypted data based on the trained neural network generating result data communicated to the user, para 23-24 and 65-66, for example). It would have been obvious to one of ordinary skill in the art before the effective filling date of the invention to include Gilad’s teaching into Rane’s invention given the benefit of scalable and efficient computation.
The data structure provided to the machine learning model would reasonably meet the limitation of the at least one machine learning data structure and a skilled in the art would readily appreciate that computing system offer their functionalities by processors executing instruction stored in memory.
Furthermore, in light of not clearly defined term “trusted environment” various interpretation of the term could be satisfied by the prior art, e.g., the environment that is executed on the machine rather than multiple machines. Additionally, even if a particular limitation would have been required, given no specific functional language farther limiting the “trusted environment”, a skilled in the art would readily appreciate that there are only two obvious environment choices within which to operate on data: trusted or non-trusted environment, selection of any of them offering the predictable benefit of customization.
Rane in view of Gilad teaches a neural network including neural network weights comprising a weight vector for neural network analytics (neural network refer to computation models with the capacity for machine learning, neurons in adjacent layers can be connected with edges, where weights are associated and learned in training phase, see Gilad’s para 1 and 21-24, for example, see the motivation to combine above) and the at least one machine learning data structure as discussed above.
Rane as modified does not teach the structure being generated based on a trained machine learning model. However, Mitchell suggests such solution (after the learning phase, the machine learning model can be used to generate search queries). It would have been obvious to one of ordinary skill in the art before the effective filling date of the invention to include known solution as taught by Michell into Rane as modified invention given the predicable benefit of generating requests.
As per claim 6 and 16, Official Notice is taken that implement the server as a cloud-based resource to be accessed by a plurality of users would have been old and well-known variant in the art at the time the application was filed offering the predictable result of scalability and customization.
Claims 3 and 13 remain rejected under 35 U.S.C. 103 as being unpatentable over Rane (USPUB 20160182222) in view of Gilad (USPUB 20160350648) and Mitchell (USPUB 20190266282), and further in view Vercauteren (Junfeng Van Vercauteren, “Somewhat Practical Fully Homomorphic Encryption”, IACR Cryptol. ePrint Arch. 2012).
Rane as modified teaches fully homomorphic encryption scheme as discussed above.
Rane does not expressly teach the homomorphie encryption including at least one of a Brakerski/Fan-Vercauteren and a Cheon-Kim-Kim-Song cryptosystem. However, having any particular system such as systems known to one of ordinary skill in the art before the effective filling date (see Vercauteren, for example) and required by the claim language would have been an obvious variant while offering the predictable benefit of customization.
Claims 10 and 20 remain rejected under 35 U.S.C. 103 as being unpatentable over Rane (USPUB 20160182222) in view of Gilad (USPUB 20160350648) and Mitchell (USPUB 20190266282), and further in view Kelleher (Kelleher et al., “Fundamentals of Machine Learning for Predictive Data Analytics, The MIT Press, Cambridge, Massachusetts, Information Based Learning Chapter, 6/2015).
Rane as modified teaches the machine learning data structure for decision analytics as discussed above.
Rane does not but Kelleher teaches the machine learning data structure comprising a tree of features and splits for a decision tree analytic (Although Chapter 4.2 expressly discusses the implementation of the decision tree features and splits implementation of machine learning analytics, applicant should review the entire text for further appreciation of the context of the entire prior art) offering the predictable benefit of improved prediction accuracy.
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 extension fee 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 date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Peter Poltorak whose telephone number is (571) 272-3840. The examiner can normally be reached Monday through Thursday from 9:00 a.m. to 5:00 p.m.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jeffrey Pwu can be reached on (571) 272-6798. The fax phone number for the organization where this application or proceeding is assigned is (571) 273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/PIOTR POLTORAK/ Primary Examiner, Art Unit 2433
Prosecution Insights