Promising Approaches to Improve the Security of Artificial Intelligence Systems

The role of cybersecurity in creating reliable and trustworthy artificial intelligence systems is especially important, as attackers conduct more and more diverse and large-scale computer attacks on artificial intelligence (AI) and constantly look for new vulnerabilities. The article presents an analysis of the threats of AI and ways to protect machine learning models. To fulfill the requirement of adaptation to security threats, the authors propose an approach of adaptive cyberimmune defense of AI.