Adversarial Attacks against ML Agents
Workshop AICA IWG Virtual Technical Workshop
Virtual
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?