Attacks on the neural network and defense methods
This work addresses security vulnerabilities in audio-based neural networks, but it appears incremental as it applies existing attack and defense methods to a specific domain.
The paper tackles adversarial attacks on neural networks trained on audio data, exploring methods like FGSM, PGD, CW, and data poisoning, and evaluates defenses using Art-IBM and advertorch libraries, presenting accuracy metrics for these scenarios.
This article will discuss the use of attacks on a neural network trained on audio data, as well as possible methods of protection against these attacks. FGSM, PGD and CW attacks, as well as data poisoning, will be considered. Within the framework of protection, Art-IBM and advertorch libraries will be considered. The obtained accuracy metrics within the framework of attack applications are presented