Systems and methods for secure tokenized credentials

1. A computer system for authentication using secure tokens, the computer system comprising computer memory, and a processor configured to:

receive one or more data sets representing one or more biometric verification credentials;

extract, from the one or more data sets, a data subset representing a constrained set of features identifying an individual;

transform the constrained set of features in combination with one or more cryptographic keys to generate a digitally signed token; and

store the digitally signed token in a data storage resident on the mobile computing device;

wherein the digitally signed token, when processed by a third party computing device, indicates that one or more characteristics of the individual are verified by one or more parties associated with each cryptographic key of the one or more cryptographic keys;

wherein the one or more data sets include one or more image sets storing a facial representation and a model data architecture trained based on speech motions of the individual, the transformation includes combining the facial representation, the model data architecture, and the one or more cryptographic keys into the digitally signed token;

wherein a processor of the third party computing device:

processes the digitally signed token upon receiving a challenge response data set representative of one or more response images representing an unverified individual speaking one or more response passphrases; and

validates the challenge response data set against the facial representation and the model data architecture to establish that the one or more response passphrases satisfies an output of the model data architecture at least at a pre-defined confidence threshold value in respect of determining that the unverified individual is the individual associated with the digitally signed token.

2. The system of claim 1 ,

wherein the digitally signed token further includes features extracted from device characteristics of the mobile computing device, and the device characteristics are utilized as part of the determination that the unverified individual is the individual associated with the digitally signed token.

3. The system of claim 1 , wherein the model data architecture is adapted to iteratively update, through modifying one or more weight values associated with one or more computing nodes, one or more transfer functions configured to minimize a loss function related to a false positive by using the speech motions of the individual as positive training examples and using reference speech motions of other individuals as negative training examples, the one or more trained transfer functions used to process the validate the challenge response data set.

4. The system of claim 3 , wherein each of the one or more passphrases includes one or more passphrase phonemes; and wherein the model data architecture includes a plurality of separate transfer functions, each corresponding to a separate passphrase phoneme, and each of the separate transfer functions generates intermediate outputs which are combined together to establish the output of the model data architecture.

5. The system of claim 4 , wherein the one or more passphrases includes a plurality of passphrases, and the third party computing device randomly or pseudo-randomly selects which of the plurality of passphrases to present as a challenge string.

6. The system of claim 3 , wherein the one or more image sets include depth values, and the speech motions of the individual are extracted from the one or more image data sets based a feature set including at least on at least one of: lateral and medial position coordinates of both eyes, lateral-position coordinates of lips, a forehead curvature, distances between an ear and the eyes, or a height of nose.

7. The system of claim 6 , wherein the constrained set of features is represented as a floating point latent vector extracted from raw image data, the floating point vector generated from an encoder neural network adapted to learn a compression of the raw image data into the floating point latent vector defined by the feature set representing speech motions of the individual.

8. The system of claim 7 , wherein the encoder neural network is a trained 2D CNN adapted to generate pixel segments on at least one of eyes, nose, and lip corners, trained to recognize people to output the floating point latent vector.

9. The system of claim 8 , wherein the digitally signed token includes a plurality of floating point latent vectors, each of the floating point latent vectors trained based on a different mix of feature sets representing speech motions of the individual.

10. The system of claim 1 , wherein the third party computing device is able to access, or stores on local memory, one or more public keys corresponding to the one or more cryptographic keys, and during processing of the digitally signed token, validates that the digitally signed token is signed by at least one of the one or more public keys.

11. A method for authentication using secure tokens, the method comprising:

receiving one or more data sets representing one or more biometric verification credentials;

extracting, from the one or more data sets, a data subset representing a constrained set of features identifying an individual;

transforming the constrained set of features in combination with one or more cryptographic keys to generate a digitally signed token; and

storing the digitally signed token in a data storage resident on the mobile computing device;

wherein the digitally signed token, when processed by a third party computing device, indicates that one or more characteristics of the individual are verified by one or more parties associated with each cryptographic key of the one or more cryptographic keys;

wherein the one or more data sets include one or more image sets storing a facial representation and a model data architecture trained based on speech motions of the individual, the transformation includes combining the facial representation, the model data architecture, and the one or more cryptographic keys into the digitally signed token;

wherein the third party computing device is configured for:

processing the digitally signed token upon receiving a challenge response data set representative of one or more response images representing an unverified individual speaking one or more response passphrases; and

validating the challenge response data set against the facial representation and the model data architecture to establish that the one or more response passphrases satisfies an output of the model data architecture at least at a pre-defined confidence threshold value in respect of determining that the unverified individual is the individual associated with the digitally signed token.

12. The method of claim 11 ,

wherein the digitally signed token further includes features extracted from device characteristics of the mobile computing device, and the device characteristics are utilized as part of the determination that the unverified individual is the individual associated with the digitally signed token.

13. The method of claim 11 , wherein the model data architecture is adapted to iteratively update, through modifying one or more weight values associated with one or more computing nodes, one or more transfer functions configured to minimize a loss function related to a false positive by using the speech motions of the individual as positive training examples and using reference speech motions of other individuals as negative training examples, the one or more trained transfer functions used to process the validate the challenge response data set.

14. The method of claim 13 , wherein each of the one or more passphrases includes one or more passphrase phonemes; and wherein the model data architecture includes a plurality of separate transfer functions, each corresponding to a separate passphrase phoneme, and each of the separate transfer functions generates intermediate outputs which are combined together to establish the output of the model data architecture.

15. The method of claim 14 , wherein the one or more passphrases includes a plurality of passphrases, and the third party computing device randomly or pseudo-randomly selects which of the plurality of passphrases to present as a challenge string.

16. The method of claim 13 , wherein the one or more image sets include depth values, and the speech motions of the individual are extracted from the one or more image data sets based a feature set including at least on at least one of: lateral and medial position coordinates of both eyes, lateral-position coordinates of lips, a forehead curvature, distances between an ear and the eyes, or a height of nose.

17. The method of claim 16 , wherein the constrained set of features is represented as a floating point latent vector extracted from raw image data, the floating point vector generated from an encoder neural network adapted to learn a compression of the raw image data into the floating point latent vector defined by the feature set representing speech motions of the individual.

18. The method of claim 17 , wherein the encoder neural network is a trained 2D CNN adapted to generate pixel segments on at least one of eyes, nose, and lip corners, trained to recognize people to output the floating point latent vector.

19. The method of claim 18 , wherein the digitally signed token includes a plurality of floating point latent vectors, each of the floating point latent vectors trained based on a different mix of feature sets representing speech motions of the individual.

20. The method of claim 11 , wherein the third party computing device is able to access, or stores on local memory, one or more public keys corresponding to the one or more cryptographic keys, and during processing of the digitally signed token, validates that the digitally signed token is signed by at least one of the one or more public keys.

21. A non-transitory computer readable medium storing machine interpretable instructions, which when executed by a processor, cause the processor to perform a method for authentication using secure tokens, the method comprising:

receiving one or more data sets representing one or more biometric verification credentials;

extracting, from the one or more data sets, a data subset representing a constrained set of features identifying an individual;

transforming the constrained set of features in combination with one or more cryptographic keys to generate a digitally signed token; and

storing the digitally signed token in a data storage resident on the mobile computing device;

wherein the digitally signed token, when processed by a third party computing device, indicates that one or more characteristics of the individual are verified by one or more parties associated with each cryptographic key of the one or more cryptographic keys;

wherein the one or more data sets include one or more image sets storing a facial representation and a model data architecture trained based on speech motions of the individual, the transformation includes combining the facial representation, the model data architecture, and the one or more cryptographic keys into the digitally signed token;

wherein the third party computing device is configured for:

processing the digitally signed token upon receiving a challenge response data set representative of one or more response images representing an unverified individual speaking one or more response passphrases; and

validating the challenge response data set against the facial representation and the model data architecture to establish that the one or more response passphrases satisfies an output of the model data architecture at least at a pre-defined confidence threshold value in respect of determining that the unverified individual is the individual associated with the digitally signed token.

22. A computer system for verification of secure tokens, the computer system comprising computer memory, and a processor configured to:

receive, from a portable computing device, a digitally signed token including a set of transformed data based at least on a constrained set of features in combination with one or more cryptographic keys, the transformed data including a model data architecture trained based on speech motions of an individual in relation to one or more passphrases;

receive a challenge response data set representative of one or more response images representing an unverified individual speaking at least one passphrase of the one or more response passphrases;

validate the challenge response data set against the model data architecture to establish that the one or more response passphrases satisfies an output of the model data architecture at least at a pre-defined confidence threshold value in respect of determining that the unverified individual is the individual associated with the digitally signed token;

validate that the digitally signed token includes at least one cryptographic signature from an approved issuer party computing device; and

upon successful validation of the challenge response data set and the at least one cryptographic signature, provision access to a controlled resource.

23. A method for verification of secure tokens, the method comprising:

receiving, from a portable computing device, a digitally signed token including a set of transformed data based at least on a constrained set of features in combination with one or more cryptographic keys, the transformed data including a model data architecture trained based on speech motions of an individual in relation to one or more passphrases;

receiving a challenge response data set representative of one or more response images representing an unverified individual speaking at least one passphrase of the one or more response passphrases;

validating the challenge response data set against the model data architecture to establish that the one or more response passphrases satisfies an output of the model data architecture at least at a pre-defined confidence threshold value in respect of determining that the unverified individual is the individual associated with the digitally signed token;

validating that the digitally signed token includes at least one cryptographic signature from an approved issuer party computing device; and

upon successful validation of the challenge response data set and the at least one cryptographic signature, provisioning access to a controlled resource.