Method and system for privacy preserving biometric authentication

1. A method for biometric authentication, implemented in a computer comprising a processor, memory accessible by the processor, and computer program instructions stored in the memory and executable by the processor, the method comprising:

storing a plurality of encrypted trained weights of a neural network classifier, wherein the weights have been trained using biometric information representing at least one biometric feature of a person, and wherein the trained weights are encrypted using homomorphic encryption, multi-party computation, or a combination of the two;

receiving encrypted biometric information obtained by sampling at least one biometric feature of the person and encrypting the sampled biometric feature;

obtaining an match-score using the encrypted trained neural network classifier, the match-score indicating a probability that the received encrypted biometric information matches the stored encrypted biometric information, wherein obtaining the match-score comprises obtaining an encrypted match-score using the encrypted trained neural network classifier, transmitting the encrypted match-score to a client device, receiving an unencrypted match-score from the client device, determining that the client correctly decrypted the match-score using a zero-knowledge proof, and comparing the unencrypted match-score with the threshold; and

authenticating the person when the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.

2. The method of claim 1 , wherein the encrypted biometric information is received from a client device communicatively connected to a biometric information acquisition device.

3. The method of claim 2 , wherein the decryption is performed at the client device using a private key.

4. The method of claim 1 , wherein some of the layers of the neural network classifier were trained using publicly-available non-private biometric information, and some layers of the neural network classifier were re-trained using private biometric information of the person.

5. The system of claim 1 , wherein the encrypted biometric information is received from a client device communicatively connected to a biometric information acquisition device.

6. The system of claim 5 , wherein the decryption is performed at the client device using a private key.

7. A system for biometric authentication, the system comprising a processor, memory accessible by the processor, and computer program instructions stored in the memory and executable by the processor to perform:

storing a plurality of encrypted trained weights of a neural network classifier, wherein the weights have been trained using biometric information representing at least one biometric feature of a person, and wherein the trained weights are encrypted using homomorphic encryption, multi-party computation, or a combination of the two;

receiving encrypted biometric information obtained by sampling at least one biometric feature of the person and encrypting the sampled biometric feature;

obtaining an match-score using the encrypted trained neural network classifier, the match-score indicating a probability that the received encrypted biometric information matches the stored encrypted biometric information, wherein obtaining the match-score comprises obtaining an encrypted match-score using the encrypted trained neural network classifier, transmitting the encrypted match-score to a client device, receiving an unencrypted match-score from the client device, determining that the client correctly decrypted the match-score using a zero-knowledge proof, and comparing the unencrypted match-score with the threshold; and

authenticating the person when the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.

8. The system of claim 7 , wherein some of the layers of the neural network classifier were trained using publicly-available non-private biometric information, and some layers of the neural network classifier were re-trained using private biometric information of the person.

9. A computer program product for biometric authentication, the computer program product comprising a non-transitory computer readable storage having program instructions embodied therewith, the program instructions executable by a computer, to cause the computer to perform a method comprising:

storing a plurality of encrypted trained weights of a neural network classifier, wherein the weights have been trained using biometric information representing at least one biometric feature of a person, and wherein the trained weights are encrypted using homomorphic encryption, multi-party computation, or a combination of the two;

receiving encrypted biometric information obtained by sampling at least one biometric feature of the person and encrypting the sampled biometric feature;

obtaining an match-score using the encrypted trained neural network classifier, the match-score indicating a probability that the received encrypted biometric information matches the stored encrypted biometric information, wherein obtaining the match-score comprises obtaining an encrypted match-score using the encrypted trained neural network classifier, transmitting the encrypted match-score to a client device, receiving an unencrypted match-score from the client device, determining that the client correctly decrypted the match-score using a zero-knowledge proof, and comparing the unencrypted match-score with the threshold; and

authenticating the person when the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.

10. The computer program product of claim 9 , wherein the encrypted biometric information is received from a client device communicatively connected to a biometric information acquisition device.

11. The computer program product of claim 10 , wherein the decryption is performed at the client device using a private key.

12. The computer program product of claim 9 , wherein some of the layers of the neural network classifier were trained using publicly-available non-private biometric information, and some layers of the neural network classifier were re-trained using private biometric information of the person.

13. A method for biometric authentication, implemented in a computer comprising a processor, memory accessible by the processor, and computer program instructions stored in the memory and executable by the processor, the method comprising:

storing a plurality of encrypted trained weights of a neural network classifier, wherein the weights have been trained using biometric information representing at least one biometric feature of a person, and wherein the trained weights are encrypted using homomorphic encryption, multi-party computation, or a combination of the two;

receiving encrypted biometric information obtained by sampling at least one biometric feature of the person and encrypting the sampled biometric feature;

obtaining an match-score using the encrypted trained neural network classifier, the match-score indicating a probability that the received encrypted biometric information matches the stored encrypted biometric information, wherein obtaining the match-score comprises obtaining an encrypted match-score using the encrypted trained neural network classifier, multiplying the encrypted match-score by a first encrypted secret integer, encrypting a plurality of additional secret integers, transmitting the multiplied encrypted match-score and at least some of the plurality of encrypted additional secret integers to a client device, receiving a decrypted multiplied match-score and decrypted additional secret integers from the client device, verifying the correctness of the decrypted additional secret integers, and dividing the unencrypted match-score by the first encrypted secret integer to obtain the match-score; and

authenticating the person when the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.

14. The method of claim 13 , wherein the multiplied encrypted match-score the encrypted additional secret integers are transmitted to the client device in a secret random order.

15. A system for biometric authentication, the system comprising a processor, memory accessible by the processor, and computer program instructions stored in the memory and executable by the processor to perform:

storing a plurality of encrypted trained weights of a neural network classifier, wherein the weights have been trained using biometric information representing at least one biometric feature of a person, and wherein the trained weights are encrypted using homomorphic encryption, multi-party computation, or a combination of the two;

receiving encrypted biometric information obtained by sampling at least one biometric feature of the person and encrypting the sampled biometric feature;

obtaining an match-score using the encrypted trained neural network classifier, the match-score indicating a probability that the received encrypted biometric information matches the stored encrypted biometric information, wherein obtaining the match-score comprises obtaining an encrypted match-score using the encrypted trained neural network classifier, multiplying the encrypted match-score by a first encrypted secret integer, encrypting a plurality of additional secret integers, transmitting the multiplied encrypted match-score and at least some of the plurality of encrypted additional secret integers to a client device, receiving a decrypted multiplied match-score and decrypted additional secret integers from the client device, verifying the correctness of the decrypted additional secret integers, and dividing the unencrypted match-score by the first encrypted secret integer to obtain the match-score; and

authenticating the person when the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.

16. The system of claim 15 , wherein the multiplied encrypted match-score the encrypted additional secret integers are transmitted to the client device in a secret random order.

17. A computer program product for biometric authentication, the computer program product comprising a non-transitory computer readable storage having program instructions embodied therewith, the program instructions executable by a computer, to cause the computer to perform a method comprising:

storing a plurality of encrypted trained weights of a neural network classifier, wherein the weights have been trained using biometric information representing at least one biometric feature of a person, and wherein the trained weights are encrypted using homomorphic encryption, multi-party computation, or a combination of the two;

receiving encrypted biometric information obtained by sampling at least one biometric feature of the person and encrypting the sampled biometric feature;

obtaining an match-score using the encrypted trained neural network classifier, the match-score indicating a probability that the received encrypted biometric information matches the stored encrypted biometric information, wherein obtaining the match-score comprises obtaining an encrypted match-score using the encrypted trained neural network classifier, multiplying the encrypted match-score by a first encrypted secret integer, encrypting a plurality of additional secret integers, transmitting the multiplied encrypted match-score and at least some of the plurality of encrypted additional secret integers to a client device, receiving a decrypted multiplied match-score and decrypted additional secret integers from the client device, verifying the correctness of the decrypted additional secret integers, and dividing the unencrypted match-score by the first encrypted secret integer to obtain the match-score; and

authenticating the person when the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.

18. The computer program product of claim 17 , wherein the multiplied encrypted match-score the encrypted additional secret integers are transmitted to the client device in a secret random order.