Use of AI in Attacks (M02b)
The use of AI in attacks is a theme that has received significant attention from both industry and academia, with numerous demonstrations, publications, and talks. In research, many AI-based techniques have been adapted and applied for supporting malicious operations and achieving a broad range of adversarial objectives. High excitement brought recently by Large Language Models and other forms of generative AI has led to a new wave of malicious AI applications and respective concerns. In the real world, however, the use of AI by cybercriminals and other attackers seems currently to come down to social engineering, while the other offensive AI capabilities stay mainly hypothetical. In this talk, we will discuss why claims about AI-enabled attacks can often be misunderstood, what evidence of such attacks we have at the moment, and the likely reasons why it is so limited. We will also look at several malicious AI applications and at examples of the (inevitable) defense-offense AI dual-use.