Reinforcement Learning for Hardware Security: Opportunities, Developments, and Challenges

TL;DR

This paper reviews how reinforcement learning can be applied to hardware security, especially in detecting hardware Trojans, highlighting recent developments, opportunities, and challenges in this emerging area.

Contribution

It provides an overview of RL applications in hardware security, focusing on hardware Trojan detection and discusses future opportunities and challenges.

Findings

RL agents can assist in hardware Trojan detection.

Application of RL in hardware security is promising but faces significant challenges.

Abstract

Reinforcement learning (RL) is a machine learning paradigm where an autonomous agent learns to make an optimal sequence of decisions by interacting with the underlying environment. The promise demonstrated by RL-guided workflows in unraveling electronic design automation problems has encouraged hardware security researchers to utilize autonomous RL agents in solving domain-specific problems. From the perspective of hardware security, such autonomous agents are appealing as they can generate optimal actions in an unknown adversarial environment. On the other hand, the continued globalization of the integrated circuit supply chain has forced chip fabrication to off-shore, untrustworthy entities, leading to increased concerns about the security of the hardware. Furthermore, the unknown adversarial environment and increasing design complexity make it challenging for defenders to detect subtle modifications made by attackers (a.k.a. hardware Trojans). In this brief, we outline the development of RL agents in detecting hardware Trojans, one of the most challenging hardware security problems. Additionally, we outline potential opportunities and enlist the challenges of applying RL to solve hardware security problems.