Biometric identity verification

An authentication device for providing biometric authentication of a test user at a time of authentication, the device comprising:

• a first input configured to receive, at the time of authentication, a first data set from a user device that is separable from the authentication device, the first data set derived from a biometric attribute of a reference user, wherein the first data set comprises multiple weight sets with each weight set corresponding to a respective neural network and each weight set corresponding to one element of the biometric attribute;
• a second input configured to receive, at the time of authentication, a test sample data set representative of a biometric attribute of the test user;
• a processor configured as a plurality of untrained artificial neural networks, the processor adapted to receive the first data set and program the untrained neural networks with the first data set as a set of weights for the untrained neural networks to generate trained neural networks;
• the processor being further adapted to receive the test sample data set as input to the trained neural networks from the second input;
• and
• a decision processor adapted to determine, from the outputs of the trained neural networks corresponding to the second input, a degree of correlation between the biometric attribute of the reference user and the biometric attribute of the test user and to authenticate the test user if the degree of correlation exceeds a threshold.

A method of biometric authentication of a test user in an authentication device, the method comprising:

• obtaining in the authentication device, at a time of authentication, a test sample data set representative of one or more biometric attributes of the test user, wherein prior to the time of authentication the authentication device includes a plurality of untrained artificial neural networks;
• receiving in the authentication device an encrypted first data set from a user device at the time of authentication, wherein the first data set is derived from one or more biometric attributes of a reference user and is stored in the user device, further wherein the first data set comprises multiple weight sets with each weight set corresponding to one element of the biometric attribute of the reference user;
• decrypting the first data set and using the decrypted first data set as a set of weights for the untrained neural networks in the authentication device to generate trained neural networks at the time of authentication;
• inputting the test sample data set as inputs to the trained neural networks;
• determining, from outputs of the trained neural networks' corresponding to the inputs, a degree of correlation between the one or more biometric attributes of the reference user and the one or more biometric attributes of the test user;
• and
• authenticating the test user if the degree of correlation exceeds a threshold.