AI vs. AI. Real Attacking and Defending in Cybersecurity

At the beginning of cybersecurity, we had rules, and attacks were detected, and for a time, all was good. And then we had a massive number of multimillion-dollar government-sponsored cloud-based advanced attacks, and we were lost. This talk will show and demo some of the most advanced projects in the Stratosphere Lab that apply AI to cybersecurity. AI vs. AI, attacking and defending, from reinforcement learning to LLMs, from deception to multi-agent environments. We will discuss ideas and limitations of detecting attacks, fighting errors, and why it is such a hard battle.