Reinforcement Learning for Micro-architectural Attack and Defense (Nov 14)

Speaker: Mulong Luo

Date: Nov 14, 2:15 – 3:05 pm

Abstract:

Securing modern computer systems against an ever-evolving threat like micro-architectural attacks is a significant challenge that requires innovative approaches. Recent developments in reinforcement learning (RL) have achieved unprecedented success in computer security. In this talk, I will present my research applying reinforcement learning methods for micro-architectural security. First, I will present AutoCAT, which uses reinforcement learning to explore cache side channel vulnerabilities in modern microprocessors. Second, I will present MACTA, which leverages multi-agent reinforcement learning to improve the side channel attack detection accuracy. Third, I will introduce our recent work leveraging RL agents for runtime cache side channel defenses. Last, I will discuss future directions.

Location: LOV 307 and ZOOM