Vulnerabilities of Artificial Intelligence Systems

The article examines artificial intelligence (AI) vulnerabilities in the context of cybersecurity, analyzes their resilience and provides vulnerability statistics. The classification of specific for the AI system vulnerabilities is given, in which 4 types of attacks are identified and described. The experiment was conducted on the simulation of an attack on a neural network using FGSM method in order to compare the efficiency of the neural network before and after the attack.