Exploring new scenarios of adversarial attacks on pattern recognition neural networks in the context of finding new defense methods

Download article in PDF format

Authors: Dyudyun G. D., Lapina M. A., Babenko M. G.

Annotation: Neural networks (NNs) are an effective tool for solving hardto-formalize problems, which has made them indispensable tools for solving them. However, information defense techniques in this area still lack sufficient protection, making them vulnerable to cybercriminals. This paper investigates adversarial attacks on neural networks, their characteristics, and proposes a new technique for detecting adversarial attacks.

Keywords: adversarial attacks, information security, machine learning, neural networks