Method of indirect dynamic biometric verification

Trokoz D. A.

The paper proposes a method of continuous biometric verification based on a heuristic algorithm using the mathematical apparatus of wide neural networks, genetic algorithms, and the algebra of hyperdimensional binary vectors. This technique is based on replacing the target verification problem with the task of predicting the time series of actions of the verified operator, providing continuous verification of his personality by analyzing behavioral characteristics. At the same time, for the synthesis of the verification model, only the data of the operator whose verification is necessary are used, which is a significant advantage compared to the methods of binary classification, the use of which requires the presence of alternative biometric images.

Keywords:neural network, dynamic verification, behavioral biometrics, time series forecasting, hyperdimensional vector algebra.