Biometric recognition and security system

1. A biometric recognition system, comprising:

a user system having:

a biometric sensor;

a processor; and

at least one non-transitory digital storage element coupled to the processor, the non-transitory digital storage element storing logic that when executed by the processor causes the processor to:

activate the biometric sensor to scan a finger of a user to provide at least one image; extract a plurality of features from the at least one image using an object detection convolutional neural network to detect and localize objects, the object detection convolutional neural network being a convolutional neural network trained to detect and localize objects, provide associated bounding boxes, class labels, and confidence and/or probability scores associated with the class labels generate a graph from the plurality of features, transform the graph into at least one fingerprint feature vector, hash the at least one fingerprint feature vector to create at least one transformed fingerprint feature vector, and transmit the at least one transformed fingerprint feature vector to at least one external device.

2. The biometric recognition system of claim 1 , wherein the transformed fingerprint feature vector serves as cryptographic key material for authentication of the user system and wherein the non-transitory digital storage element stores logic that when executed by the processor causes the processor to enroll the user and deploy the at least one transformed fingerprint feature vector to a plurality of user devices of the user such that the plurality of devices is operable to locally authenticate the user.

3. The biometric recognition system of claim 1 , wherein the transformed fingerprint feature vector serves as cryptographic key material for identification of the user.

4. The biometric recognition system of claim 1 , wherein the plurality of features extracted from the at least one image includes minutiae of a fingerprint.

5. The biometric recognition system of claim 1 , wherein extracting the plurality of features using the object detection convolutional neural network further includes the step of:

receiving the at least one image and detecting at least one minutiae within the at least one image from at least one feature map.

6. The biometric recognition system of claim 5 , wherein extracting the plurality of features to create at least one fingerprint feature vector using the object detection convolutional neural network further includes the step of:

receiving the at least one image and detecting at least singular point within the at least one image from the at least one feature map.

7. The biometric recognition system of claim 5 , wherein extracting the plurality of features to create at least one fingerprint feature vector using the object detection convolutional neural network further includes the step of:

extracting at least one feature from the image relating to the location of an object localized by the object detection convolutional neural network.

8. The biometric recognition system of claim 5 , wherein extracting the plurality of features to create at least one fingerprint feature vector using the object detection convolutional neural network further includes the step of:

providing, using the extracted plurality of features, the graph having at least one node, each node represented by minutiae location, minutiae orientation, or type of minutiae; with each node representing detected features and relations between each node modeled as edges such that edges denote connectors; and wherein edges are described by direction and number of ridge crossings.

9. The biometric recognition system of claim 8 , wherein entirety of the graph includes a plurality of local regions with each local region characterized by local relational attributes.

10. The biometric recognition system of claim 8 , wherein extracting the plurality of features to create at least one fingerprint feature vector using the object detection convolutional neural network further includes the step of:

creating, from the graph, at least one N×N square adjacency matrix, wherein N is an amount of nodes in the graph and values of individual elements within the at least one N×N square adjacency matrix reflect presence of at least one edge, the at least one N×N square adjacency matrix having relative location information of at least one salient feature in the at least one image and wherein the at least one N×N square adjacency matrix is padded to accommodate a maximum number of expected minutiae.

11. The biometric recognition system of claim 1 , the at least one fingerprint feature vector is compressed by an autoencoder providing a compressed representation of the fingerprint feature vector.

12. The biometric recognition system of claim 11 , wherein the autoencoder includes an encoder function configured to transform the at least one fingerprint feature vector into at least one latent representation having lower dimensionality than the at least one fingerprint feature vector being input into the autoencoder.

13. The biometric recognition system of claim 11 , wherein the compressed representation of the fingerprint feature vector is binarized and hashed.

14. The biometric recognition system of claim 1 , wherein the fingerprint feature vector is used in biometric key-signing authentication and the non-transitory digital storage element storing logic further causes the processor to:

transmit identity data from the user;

respond to a challenge from the external device transforming the challenge data using the transformed fingerprint feature vector; and

transmit the transformed challenge data to the external device.

15. The biometric recognition system of claim 1 , further comprising transforming the at least one fingerprint feature vector as an input to seed a random number generator.

16. The biometric recognition system of claim 15 , wherein the at least one fingerprint feature vector is transformed via a keypair using key derivation functions.

17. The biometric recognition system of claim 16 , wherein the transformed fingerprint feature vector is computationally irreversible.

18. A method, comprising:

activating, by a user system, a biometric sensor to scan a finger of a user to provide at least one image;

extracting, by the user system, a plurality of features from the at least one image using an object detection convolutional neural network to detect and localize objects, the object detection convolutional neural network being a convolutional neural network trained to detect and localize objects, provide associated bounding boxes, class labels, and confidence and/or probability scores associated with the class labels;

generating a graph from the plurality of features extracted by the object detection convolutional neural network;

transforming, by the user system, the graph into at least one fingerprint feature vector;

hashing the at least one fingerprint feature vector to create at least one transformed fingerprint feature vector; and,

transmitting, by the user system, the at least one transformed fingerprint feature vector to at least one external device.