Systems and methods for privacy-enabled biometric processing
A set of distance measurable encrypted feature vectors can be derived from any biometric data and/or physical or logical user behavioral data, and then using an associated deep neural network (“DNN”) on the output (i.e., biometric feature vector and/or behavioral feature vectors, etc.) an authentication system can determine matches or execute searches on encrypted data. Behavioral or biometric encrypted feature vectors can be stored and/or used in conjunction with respective classifications, or in subsequent comparisons without fear of compromising the original data. In various embodiments, the original behavioral and/or biometric data is discarded responsive to generating the encrypted vectors. In another embodiment, distance measurable or homomorphic encryption enables computations and comparisons on cypher-text without decryption of the encrypted feature vectors. Security of such privacy enabled embeddings can be increased by implementing an assurance factor (e.g., liveness) to establish a submitted credential has not been spoofed or faked.
1. A privacy-enabled authentication system comprising:
at least one processor operatively connected to a memory, the at least one processor configured to:
execute a first machine learning (“ML”) and a second ML process responsive to an authentication mode;
determine the authentication mode;
trigger one or both of the first ML process or the second ML process responsive to determining the authentication mode;
wherein the first ML process when executed by the at least one processor is configured to:
accept distance measurable encrypted feature vector and label inputs during training of one or more first classification neural networks to define a plurality of identification classes and classify distance measurable encrypted feature vector inputs as part of authentication using the one or more first classification neural networks once trained;
wherein the second ML process when executed by the at least one processor is configured to:
accept plain text biometric or behavioral inputs as input to one or more generation neural networks and output respective distance measurable encrypted feature vectors;
compare distances between distance measurable encrypted feature vectors generated by respective neural networks during authentication; and
validate identification results produced by the first and second ML processes are captured from a live submission, the validation including operations to determine liveness in multiple dimensions including at least liveness evaluation of authentication inputs of a matching type submitted to the first or second ML process, as part of the evaluation of the multiple dimensions.
2. The system of claim 1 , wherein one of the first ML process or the second ML process is configured to:
determine one or more distances between encrypted feature vectors produced by respective generation neural networks;
exclude encrypted feature vectors produced by respective generation neural networks having one or more distances exceeding a threshold distance for subsequent training processes; and
include encrypted feature vectors having distances within the threshold distance for subsequent training processes.
3. The system of claim 1 , wherein the at least one processor is configured to determine the authentication mode includes an enrollment mode for establishing a new entity for subsequent authentication.
4. The system of claim 3 , wherein the at least one processor is configured to trigger at least the second ML process responsive to determining a current authentication mode includes the enrollment mode.
5. The system of claim 3 , wherein the at least one processor is configured to trigger at least training operations of both the first and second ML processes responsive to determining that the current authentication mode includes the enrollment mode.
6. The system of claim 5 , wherein the at least one processor is configured to execute at least the second ML process to authenticate a new user until at least a period of time required for training the one or more first classification neural networks expires.
7. The system of claim 5 , wherein the at least one processor is configured to execute at least the first ML process to authenticate a new user responsive to completing training of the one or more first classification neural networks.
8. The system of claim 1 , wherein the one or more first classification neural networks comprise a deep neural network (“DNN”), wherein the DNN is configured to:
generate an array of values in response to the input of the at least one unclassified encrypted feature vector during authentication; and
determine a label or unknown result based on analyzing the generated array of values.
9. The system of claim 1 , wherein the generation neural networks comprise at least one learning network configured to accept plain text biometric as input and generate distance measurable encrypted feature vectors as output.
10. The system of claim 1 , wherein the one or more first classification neural networks are configured to return a label for identification or an unknown result, responsive to input of encrypted feature vector input.
11. The system of claim 1 , wherein the at least one processor is configured to:
determine a probability of match using the one or more first classification neural networks is below a threshold value; and
validate an unknown result output by the one or more first classification neural networks based on distance analysis of a highest probability match compared to the input feature vectors.
12. A privacy-enabled authentication system comprising:
at least one processor operatively connected to a memory, the at least one processor configured to:
execute a first machine learning (“ML”) process, wherein the first ML process when executed by the at least one processor is configured to:
validate training inputs comprising distance measurable encrypted feature vector produced by one or more generation networks;
reject any feature vector if during validation the distances between the distance measurable feature vectors produced by a respective generation network are greater than a validation threshold; and
accept the validated distance measurable encrypted feature vectors produced by the one or more generation networks and associated identification label inputs during training of one or more classification neural networks; and
execute a second machine learning (“ML”) process, wherein the second ML process when executed by the at least one processor is configured to classify distance measurable encrypted feature vector inputs as part of authentication using the one or more classification neural networks once trained;
validate results from the one or more classification neural networks are captured from a live submission, the validation of the results from the one or more classification neural networks including operations to determine liveness in multiple dimensions, including at least liveness evaluation of authentication inputs of a matching type submitted to the first or second ML process, as part of the evaluation of the multiple dimensions.
13. The system of claim 12 , wherein the system defines a validation threshold associated with the output of each generation network.
14. The system of claim 13 , wherein the system defines the validation threshold based at least in part on a percentage deviation from an identification threshold.
15. The system of claim 14 , wherein the identification threshold is established when two encrypted feature vectors produced by a respective generation network are determined to be associated with a single entity or object.
16. The system of claim 12 , wherein the at least one processor is configured to validate identification results produced by the first ML processes, the validation including operations to determine liveness in multiple dimensions including at least liveness evaluation of authentication inputs of a matching type submitted to the first ML process.
17. A computer implemented method for privacy enabled authentication, the method comprising:
determine, by at least one processor, an authentication mode;
triggering, by the at least one processor, one or both of a first machine learning (“ML”) process or a second ML process responsive to determining the authentication mode;
executing, by the at least one processor, the first ML process, wherein executing the first ML process includes:
accepting distance measurable encrypted feature vector and label inputs during training of a first classification neural network and classifying distance measurable encrypted feature vector inputs as part of authentication using the first classification neural network once trained;
executing, by the at least one processor, the second ML process, wherein executing the second ML process includes:
accepting plain text biometric inputs during training of a generation neural network to generate distance measurable encrypted feature vectors;
comparing distances between distance measurable encrypted feature vectors during authentication; and
validating, by the at least one processor, identification results produced by the first and second ML processes are captured from a live submission, wherein validating includes determining liveness in multiple dimensions including at least liveness evaluation of authentication inputs of a matching type submitted to the first or second ML process, as part of the evaluation of the multiple dimensions.
18. The method of claim 17 , further comprising:
determining one or more distances between encrypted feature vectors produced by the generation neural network;
excluding encrypted feature vectors having one or more distances exceeding a threshold distance for subsequent training processes; and
including encrypted feature vectors having distances within the threshold distance for subsequent training processes.