Adversarial Attacks against ML Agents

Workshop AICA IWG Virtual Technical Workshop


Oneliner: Addressing the resilience of AICA against adversarial ML attacks.

Autonomous Intelligent Cyber Agents – AICA in short – will rely on ML but will also be deployed in adversarial settings. Hence, they will undeniably be targeted by cyber attacks. In my talk, I will show (i) how AICA can be attacked, (ii) what are the limitations of existing work, and (iii) where should we act to improve the resilience of AICA?

Slides Video (TBD) Venue