Vulnerabilities of AI in Cybersecurity

Shahabuddin Shahid

Abstract: Artificial intelligence create new ways of attack that adversaries could take advantages of as it becomes more connected to cybersecurity systems. Two important vulnerabilities are evaluated in this paper: Model poisoning and adversarial attacks. We look at how these issues can damage overall security, dependability and integration of technologies that AI is used in, especially in cybersecurity applications. The risks of releasing AI in an adversarial environment are increasingly recognized by new researchers and developers.

Keywords: Artificial Intelligence, Cybersecurity, adversarial attacks, Model poisoning, Security Weaknesses, Machine Learning Security, Threat Detection