Systems and methods for privacy-enabled biometric processing
In one embodiment, a set of feature vectors can be derived from any biometric data, and then using a deep neural network (“DNN”) on those one-way homomorphic encryptions (i.e., each biometrics' feature vector) can determine matches or execute searches on encrypted data. Each biometrics' feature vector can then be stored and/or used in conjunction with respective classifications, for use in subsequent comparisons without fear of compromising the original biometric data. In various embodiments, the original biometric data is discarded responsive to generating the encrypted values. In another embodiment, the homomorphic encryption enables computations and comparisons on cypher text without decryption. This improves security over conventional approaches. Searching biometrics in the clear on any system, represents a significant security vulnerability. In various examples described herein, only the one-way encrypted biometric data is available on a given device. Various embodiments restrict execution to occur on encrypted biometrics for any matching or searching.
1. A privacy-enabled biometric system comprising:
at least one processor operatively connected to a memory, wherein the at least one processor when executing is configured to;
instantiate a respective member of a first operative pairing of neural networks, the first operative pairing including at least one generation neural network and at least one classification neural network;
wherein the respective member of the first operative pairing of neural networks comprises the at least one classification neural network configured to:
accept as an input distance measurable encrypted feature vectors, the distance measurable encrypted feature vectors generated as a one way encoding of plain text authentication information input to at least one first neural network, wherein the at least one classification neural network is trained on distance measurable encrypted feature vector and respective label inputs; and
predict a match to a label for identification, authentication, or to return unknown responsive to input of at least one distance measurable encrypted feature vector produced by the at least one first neural network.
2. The system of claim 1 , wherein the at least one classification neural network is configured to authenticate or identify an entity based on predicting a match to the label meeting a threshold probability.
3. The system of claim 1 , wherein the at least one processor when executing is configured to;
instantiate a second respective member of the first operative pairing of neural networks comprising the at least one generation neural network.
4. The system of claim 3 , wherein the at least one generation neural network comprises a pre-trained neural network configured to generate the one way encoding of plain text authentication information.
5. The system of claim 4 , wherein the at least one processor when executing is configured to:
instantiate a plurality of pre-trained neural networks, each pre-trained neural network paired with respective classification neural networks based on a type associated with input identification information.
6. The system of claim 1 , wherein the at least one processor when executing is configured to instantiate multiple classification neural networks, each classification neural network associated with a biometric type and a respective pre-trained neural network.
7. The system of claim 6 , wherein respective pairs of the classification neural network and the pre-trained neural network are configured to authenticate an entity based on an input of a type of authentication information.
8. The system of claim 1 , wherein the at least one processor when executing is configured to:
manage a plurality of modes of execution, including an enrollment mode configured to accept a label for identification and associated encrypted feature vectors for an entity.
9. The system of claim 1 , wherein the at least one classification neural network is configured to predict a match to an existing label or to return an unknown result based on encrypted feature vectors enrolled on the system.
10. The system of claim 1 , wherein the at least one classification neural network further comprises an input layer for accepting feature vectors of a number of dimensions, the input layer having a number of nodes at least equal to the number of dimensions of the feature vectors of the number of dimensions, a first hidden layer, a second hidden layer, and an output layer that generates an array of values that based on a respective position of the values in the array and the values at each position, determine the label or unknown.
11. A computer implemented method for privacy-enabled biometric authentication, the method comprising:
instantiating, by at least one processor, a respective member of a first operative pairing of neural networks, the first operative pairing including at least one generation neural network and at least one classification neural network, and wherein the respective member of the first operative pairing of neural networks comprises the at least one classification neural network;
accepting, by the at least one classification neural network, as an input distance measurable encrypted feature vectors, wherein the distance measurable encrypted feature vectors are generated as a one way encoding of plain text authentication information input to at least one generation neural network;
training, by at least one processor, the at least one classification neural network on distance measurable encrypted feature vector and respective label inputs; and
predicting, by the at least one classification neural network, a match to a label for identification or returning unknown responsive to input of at least one distance measurable encrypted feature vector produced by the at least one first neural network.
12. The method of claim 11 , wherein the method further comprising confirming identification or authentication based on predicting a match to the label meeting a threshold probability.
13. The method of claim 11 , further comprising instantiating the at least one generation neural network.
14. The method of claim 13 , wherein the act of instantiating the at least one generation neural network includes instantiating a first pre-trained generation neural network; and the method further comprises generating, by the first pre-trained generation neural network the one way encoding of plain text authentication information.
15. The method of claim 14 , wherein the method further comprises:
instantiating, by the at least one processor, a plurality of pre-trained generation neural networks, each pre-trained neural network paired with respective classification neural networks, and
wherein the respective classification neural networks are trained to classify the one way encodings produced by paired pre-trained neural networks.
16. The method of claim 15 , wherein the method further comprises instantiating multiple pre-trained generation neural networks for processing respective types of identification or authentication information responsive to identifying submission of multiple types of identifying information.
17. The method of claim 11 , wherein the method further comprises instantiating multiple classification neural networks, each classification neural network associated with a biometric type and a respective pre-trained generation neural network.
18. The method of claim 17 , wherein method further comprises:
instantiating respective members of respective operative pairings of the generation neural network and the pre-trained generation neural network; and authenticating or identifying, by the respective operative pairings, an entity based on an input of a type of authentication information.
19. The method of claim 11 , wherein the method further comprises triggering one of a plurality of modes of execution, including an enrollment mode configured to accept a label for identification and associated encrypted feature vectors for an entity.
20. The method of claim 11 , wherein the method further comprises predicting a match to an existing label or returning an unknown result based on encrypted feature vectors enrolled on the system.
21. The method of claim 11 , wherein the act of instantiating the classification network further comprises instantiating the ate least one classification neural network having an input layer for accepting encrypted feature vectors of a number of dimensions, the input layer having a number of nodes at least equal to the number of dimensions of the feature vectors, a first hidden layer, a second hidden layer, and an output layer that generates an array of values that based on a respective position of the values in the array and the values at each position; and the method further comprises determining the label or unknown based probabilities from or derived from the array of values.