Research Manuscript: Security of AI and AI for security
Event TypeResearch Manuscript
Hardware Security: Attack and Defense
TimeWednesday, July 13th1:30pm - 3:00pm PDT
Location3006, Level 3
DescriptionArtificial Intelligence (AI) is increasingly used for evaluating weaknesses in secure systems. On the other hand, it is important to protect AI systems against attacks that steal the IP of the neural network model and fault attacks that jeopardize the safety of the system. This session presents one paper on a defense mechanism against IP theft and one paper on a DVFS-based fault attack on DNN hardware. The session also contains a paper that uses adversarial learning to perform a profiled side-channel attack and a paper that uses reinforcement learning to detect hardware Trojans.