DETAILED ACTION
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 .
Claim Objections
Claim 18 is objected to because of the following informalities: it would appear that claim 18 should be depended on claim 17. Appropriate correction is required.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1, 6, 8-11 and 17-20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Pandaze et al. (“Poster: Privacy-Preserving Neural Network with Functional Encryption”, hereinafter Pandaze).
Re. claim 1, Pandaze discloses a process to analyze data in an artificial neural network comprising: receiving the data into the artificial neural network (Pandaze discloses machine learning in computer vision [Page 1, I Introduction] the input layers consists of an encrypted vector of N elements. If the input data fed to the neural network is not vectorized, it must be vectorized. If an input is a grayscale image, it should be flattened to a vector [Page 1, Section III The proposed method] Fig. 1 shows a computer implementing machine learning);
functionally encrypting the data at a plurality of inner input nodes in the artificial neural network (Pandaze discloses the first hidden layer, we used our secure activation function called SecureReLU. SecureRelu function takes the input function encrypted vector X encrypted using FHIPE or IPFE, weight matrix W, and a bias term [Page 1, Section III The proposed method]. IPFE or FHIPE is mainly used for computing products between weight matrix and vector X. Here, X is encrypted with mpk (master public key), and access to the result, i.e., the product of W and X, is given by using the strength of FE. We detail the working of encryption and SecureReLU function in Algorithm 1 and Algorithm 2, respectively. In the process of the first activation function in the forward propagation, SecureReLU is used. In the later stages, the processes may occur similar to regular neural networks. Similarly, the backpropagation is made secure by computing the inner products using FE [Page 2, Section III The proposed method] Fig. 1);
processing the functionally encrypted data in the artificial neural network (Pandaze discloses the first hidden layer, we used our secure activation function called SecureReLU. SecureRelu function takes the input function encrypted vector X encrypted using FHIPE or IPFE, weight matrix W, and a bias term [Page 1, Section III The proposed method]. IPFE or FHIPE is mainly used for computing products between weight matrix and vector X. Here, X is encrypted with mpk (master public key), and access to the result, i.e., the product of W and X, is given by using the strength of FE. We detail the working of encryption and SecureReLU function in Algorithm 1 and Algorithm 2, respectively. In the process of the first activation function in the forward propagation, SecureReLU is used. In the later stages, the processes may occur similar to regular neural networks. Similarly, the backpropagation is made secure by computing the inner products using FE [Page 2, Section III The proposed method] Fig. 1); and generating an output from the processing of the functionally encrypted data in the artificial neural network (Pandaze discloses SecureReLU returns the results equivalent to ReLU(WTX+b), which usually happens in the activation functions in modern neural networks. So, the SecureReLU activation function looks like: A=ReLU(skfe(W) * Encrypt(X) + b ) [Page 2, Section III The proposed method] Fig. 1, Algorithm 1 and Algorithm 2 discloses output).
Re. claim 6, Pandaze discloses the process of claim 1,wherein the data comprise medical image data (Pandaze discloses the input layers consists of an encrypted vector of N elements. If the input data fed to the neural network is not vectorized, it must be vectorized. If an input is a grayscale image, it should be flattened to a vector [Page 1, Section III The proposed method] Fig. 1 grayscale image as medical image data).
Re. claim 8, Pandaze disclose the process of claim 1, wherein the plurality of inner nodes comprises a plurality of nodes in a first input layer of the artificial neural network (Pandaze discloses the first hidden layer, we used our secure activation function called SecureReLU. SecureRelu function takes the input function encrypted vector X encrypted using FHIPE or IPFE, weight matrix W, and a bias term [Page 1, Section III The proposed method]. IPFE or FHIPE is mainly used for computing products between weight matrix and vector X. Here, X is encrypted with mpk (master public key), and access to the result, i.e., the product of W and X, is given by using the strength of FE. We detail the working of encryption and SecureReLU function in Algorithm 1 and Algorithm 2, respectively. In the process of the first activation function in the forward propagation, SecureReLU is used. In the later stages, the processes may occur similar to regular neural networks. Similarly, the backpropagation is made secure by computing the inner products using FE [Page 2, Section III The proposed method] Fig. 1).
Re. claim 9, Pandaze disclose the process of claim 8, wherein the plurality of nodes in the first input layer of the artificial neural network comprises all the nodes in the first input layer of the artificial neural network (Pandaze discloses the first hidden layer, we used our secure activation function called SecureReLU. SecureRelu function takes the input function encrypted vector X encrypted using FHIPE or IPFE, weight matrix W, and a bias term [Page 1, Section III The proposed method]. IPFE or FHIPE is mainly used for computing products between weight matrix and vector X. Here, X is encrypted with mpk (master public key), and access to the result, i.e., the product of W and X, is given by using the strength of FE. We detail the working of encryption and SecureReLU function in Algorithm 1 and Algorithm 2, respectively. In the process of the first activation function in the forward propagation, SecureReLU is used. In the later stages, the processes may occur similar to regular neural networks. Similarly, the backpropagation is made secure by computing the inner products using FE [Page 2, Section III The proposed method] Fig. 1).
Re. claim 10, Pandaze discloses the process of claim 1,wherein the data comprise training data and input data (Pandaze discloses a complete PPML pipline that successfully performs training and inference on the encrypted data using functional encryption [Page 1, abstract]).
Re. claim 11, claim 11 is rejected with the same rationale as applied in claim 1 above.
Re. claim 17, rejection of claim 11 is included and claim 17 is rejected with the same rationale as applied in claim 8 above.
Re. claim 18, rejection of claim 11 is included and claim 17 is rejected with the same rationale as applied in claim 9 above.
Re. claim 19, rejection of claim 11 is included and claim 19 is rejected with the same rationale as applied in claim 10 above.
Re. claim 20, claim 20 is rejected with the same rationale as applied in claim 1 above.
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 2-5 and 12-15 are rejected under 35 U.S.C. 103 as being unpatentable over Pandaze et al. (“Poster: Privacy-Preserving Neural Network with Functional Encryption”, hereinafter Pandaze) in view of Manevich et al. (US 20220278845, hereinafter MAnevich).
Re. claim 2, Pandaze discloses the process of claim 1,wherein the functionally encrypting the data comprises: generating public parameters and a master secret key using a security parameter; generating functional secret keys for an input function using the master secret key; and functionally encrypting the data into ciphertext using the master secret key (Manevich teaches creates a master secret key sk and its public key pk [0057]. Encrypt(pk; u): Encrypts a vector u and yields ciphertext c [0058]. Encrypt the vector u using the public key (pk) of the regulator generated in FIG. 4B to generate ciphertext (c). Generate a fee estimation (x) by performing a functional encryption via inner product (e.g., x=<v, u>) based off the ciphertext (c) and the functional encryption key (sk.sub.v) [0094]).
Therefore, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to modify the method and system disclosed by Pandaze to include generating public parameters and a master secret key using a security parameter; generating functional secret keys for an input function using the master secret key; and functionally encrypting the data into ciphertext using the master secret key as disclosed by Manevich. One of ordinary skill in the art would have been motivated for the purpose of to prove that the correct functional encryption key was used and to encrypt data (Manevich [0054]).
Re. claim 3, Pandaze-Manevich teach the process of claim 2, wherein the functionally encrypted ciphertext generated by the functional encryption of the data at the plurality of inner input nodes in the artificial neural network generates ciphertext for the data and the input function of the artificial neural network, thereby protecting the data and the input function of the artificial neural network (Manevich teaches creates a master secret key sk and its public key pk [0057]. Encrypt(pk; u): Encrypts a vector u and yields ciphertext c [0058]. Encrypt the vector u using the public key (pk) of the regulator generated in FIG. 4B to generate ciphertext (c). Generate a fee estimation (x) by performing a functional encryption via inner product (e.g., x=<v, u>) based off the ciphertext (c) and the functional encryption key (sk.sub.v) [0094][0150] Fig. 8A).
Therefore, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to modify the method and system disclosed by Pandaze to include generating public parameters and a master secret key using a security parameter; generating functional secret keys for an input function using the master secret key; and functionally encrypting the data into ciphertext using the master secret key as disclosed by Manevich. One of ordinary skill in the art would have been motivated for the purpose of to prove that the correct functional encryption key was used and to encrypt data (Manevich [0054]).
Re. claim 4, Pandaze-Manevich teach the process of claim 3,wherein the input function comprises a weight of a node in the artificial neural network multiplied by the data (Pandaze disclose SecureReLU function takes the input encrypted vector X encrypted using FHIPE or IPFE, weight matrix W, and a bias term. SecureReLU returns the results equivalent to ReLU(WTX+b), which usually happens in the activation functions in modern neural networks [Page 1, Section III The proposed method]. computing products between weight matrix and vector X [Page 2, Section III The proposed method]).
Re. claim 5, Pandaze-Manevich teach the process of claim 2, comprising decrypting the functionally encrypted cyphertext using the functional secret key and the public parameters (Manevich teaches the functional encryption key sk.sub.v, and the ciphertext c into a decrypt operation [0054]. Decrypt(pk; c; skv): Outputs <v, u> [0060]).
Therefore, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to modify the method and system disclosed by Pandaze to include decrypting the functionally encrypted cyphertext using the functional secret key and the public parameters as disclosed by Manevich. One of ordinary skill in the art would have been motivated for the purpose of to prove that the correct functional encryption key was used (Manevich [0054]).
Re. claim 12, rejection of claim 11 is included and claim 12 is rejected with the same rationale as applied in claim 2 above.
Re. claim 13, rejection of claim 12 is included and claim 13 is rejected with the same rationale as applied in claim 3 above.
Re. claim 14, rejection of claim 13 is included and claim 14 is rejected with the same rationale as applied in claim 4 above.
Re. claim 15, rejection of claim 12 is included and claim 15 is rejected with the same rationale as applied in claim 5 above.
Claims 7 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Pandaze et al. (“Poster: Privacy-Preserving Neural Network with Functional Encryption”, hereinafter Pandaze) in view of Dey et al. (EP 4181004 hereinafter Dey).
Re. claim 7, Pandaze discloses the process of claim 6, wherein the output from the artificial neural network comprises a medical analysis or a medical prediction (Dey teaches the PPMMD 110 may be utilized to perform machine learning processes in a distributed or federated manner that enables the hospitals to gain insights into patient conditions (e.g., medical conditions, symptoms of medical conditions, efficacy of treatments for the medical conditions, and the like) based on medical data while preserving the privacy of patient data utilized by the machine learning models. In this manner, more rapid understanding of medical conditions may be achieved via machine learning techniques and the accuracy and predictive capabilities of the machine learnings models may be improved through the use of data sets across multiple hospitals or hospital systems [0015]. he machine learning models may be utilized to perform diagnostic or predictive analysis of user data (e.g., the above described patient data), such as to diagnose a patient condition, predict progression of medical conditions, effectiveness of treatments, other diagnostic or predictive tasks, or combinations thereof [0063]).
Therefore, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to modify the method and system disclosed by Pandaze to include wherein the output from the artificial neural network comprises a medical analysis or a medical prediction as disclosed by Dey. One of ordinary skill in the art would have been motivated for the purpose of evaluate patients with more accuracy (Dey [0063]).
Re. claim 16, rejection of claim 11 is included and claim 16 is rejected with the same rationale as applied in claim 7 above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Tomida (US 20220140998) discloses encryption and decryption by a multi-input inner product functional encryption having a function hiding property includes a setup unit configured to generate, taking a vector length m and the number of arguments μ of an inner product function as input, a master secret key msk and a public parameter pp by using a setup algorithm of a single-input inner product functional encryption having a predetermined characteristic and having a function hiding property.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KEVIN A AYALA whose telephone number is (571)270-3912. The examiner can normally be reached Monday-Thursday 8AM-5PM; Friday: Variable EST.
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, Jorge Ortiz-Criado can be reached at 571-272-7624. 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.
/KEVIN AYALA/Primary Examiner, Art Unit 2496
Prosecution Insights