An Overview of Machine Learning Applications in Hardware Security

Abstract

this study explores the uses of machine learning (ML) in the field of hardware security, in particular, two applications areas are considered, namely, hardware Trojan (HT) and IC counterfeits. These examples demonstrate how ML algorithms can be employed as a defense mechanism to detect forged or tampered-with circuits. Our analysis shows that the ML detection accuracy still has not reached 100%. The selection and size of the feature vectors greatly affect the performance of the learning models, however, increasing the number of features or their size can lead to large overheads. Therefore, a thorough analysis is required to only select the appropriate- ate several relevant features that significantly contribute to the accuracy of ML models. The study also highlighted the need for a more robust deployment of ML algorithms to enhance their resilience to adversarial attacks.