5. Machine Learning in Hardware Security

The ever-increasing demand for higher computing capabilities constantly pushes the development of the semiconductor industry and is making modern hardware an extremely complicated artifact. However, with the increased complexity of today’s hardware systems, their security is also challenged by various vulnerabilities coming from different perspectives. Hardware-related security, as an emerging research area, has been gaining more attention during the past decades. As a result, more novel attacks and corresponding countermeasures are being proposed almost every day. The complexity of modern hardware systems makes many conventional methodologies reaching the limit of their analytical capabilities, and new powerful methods and tools are in urgent need to study hardware security problems. Thanks to the significant development of machine learning, numerous advanced analytical methodologies and tools become directly available and applicable to hardware security research, which greatly enhances the ability of both hardware designers and attackers. To present readers the important role played by machine learning in today’s hardware security, this chapter presents the application of machine learning in different hardware security areas, such as IP protection, Trojan detection, side-channel analysis/attacks, hardware security primitives, and architectural vulnerabilities, and highlights future research directions.