METHOD OF PROTECTING FINGERPRINTS IN AN INTELLIGENT DECISION-MAKING SYSTEM BASED ON CONVULSIVE NEURAL NETWORKS

Abstract

The article investigates the problem of detecting fingerprint "liveness" as an important element of ensuring cybersecurity of modern biometric authentication systems. It is substantiated that the widespread use of biometrics in financial services, corporate networks, mobile applications and e-government systems is accompanied by an increase in the risks of spoofing attacks, during which attackers use artificial or modified fingerprints to bypass identification mechanisms. An approach to solving this problem is proposed based on convolutional neural networks, capable of automatically identifying discriminative features of images, in particular textural features, papillary line structure violations and artifacts characteristic of counterfeit samples.

Experimental studies were performed using the SocoFing dataset with a constant model configuration and training parameters. The results obtained demonstrate the high efficiency of the approach: the classification accuracy is 98.964%, the false acceptance rate (FAR) is 0.215%, and the false rejection rate (FRR) is 7.251%. A low FAR value indicates the model’s ability to minimize the risk of unauthorized access, which is critically important for systems with increased security requirements. At the same time, an increased level of FRR indicates the need for further optimization to ensure a balance between security and usability.

It is concluded that models based on deep learning can be effectively used as an additional layer of protection in biometric systems, especially in the context of multi-factor authentication and the concept of zero trust. Directions for further research are outlined, in particular, the use of more diverse data sets, increasing resistance to new types of attacks, and integration with other cyber defense mechanisms.