Hardware-assisted cybersecurity countermeasures by employing applications’ low-level features collected from Hardware Performance Counters (HPCs) registers have emerged as a promising solution to address the inefficiency of traditional software-based methods. This work proposes an accurate and cost-efficient hardware accelerated machine learning-based solutions for securing emerging edge devices against malware using processors’ HPCs data.
In this paper, we proposed \textit{OptiEdge}, an ML-based hardware-assisted resource and timing estimation tool that can reduce the design space exploration for edge devices' design. We also comprehensively explored the suitability of applying various types of machine learning classifiers for hardware-assisted malware detection by precisely comparing them in terms of detection accuracy, throughput, and hardware overheads.