Adaptive machine learning for manufactured IC image analysis

• Main Author: Tee, Yee Yang
• Other Authors: Gwee Bah Hwee
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Engineering; Machine learning; Image analysis; Hardware assurance; Image synthesis; Domain adaptation; Adversarial learning
• Online Access: https://hdl.handle.net/10356/181526
• Institution: Nanyang Technological University

The hardware assurance of integrated circuits (ICs), which is concerned with ensuring the security and integrity of ICs, is of paramount importance today given the ubiquity and necessity of semiconductor devices worldwide. Circuit extraction via the analysis of integrated circuit (IC) images is one of the most reliable approaches to hardware assurance. Deep learning techniques have been demonstrated to achieve higher accuracy and speed for IC image analysis as compared to conventional methods. However, massive amounts of high-quality training data are required which can be extremely time-consuming and costly to obtain. In this thesis, adaptive machine learning techniques encompassing image synthesis, adversarial supervision, and domain adaptation were proposed to reduce the dependence on labelled training data and improve the accuracy of IC image analysis. The outcome of this thesis work will greatly enhance the efficiency and productivity of hardware assurance processes.