Keynotes

Abstract

Over the last fifteen years, hardware security and trust has evolved into a major new area of research at the intersection of semiconductor manufacturing, VLSI design and test, computer-aided design, architecture and system security. During the same period, machine learning has experience a major revival in interest and has flourished from a nearly forgotten area to the talk of the town. In this presentation, we will first briefly review various machine learning-based solutions which have been developed to address a number of concerns in hardware security and trust, including hardware Trojan detection, counterfeit IC identification, provenance attestation, hardware-based malware detection, side-channel attacks, PUF modeling, etc. Then, we will examine the key attributes of these problems which make them amenable to machine learning-based solutions and we will discuss the potential and the fundamental limitations of such approaches. Lastly, we will ponder the role of and necessity for advanced contemporary machine learning methods in the context of hardware security and we will conclude with suggestions for avoiding common pitfalls when employing such methods.