Power MOSFETs Identification Using Machine Learning on Characteristic Curves

摘要

Device identification of power MOSFETs is important to improve the security of critical systems. We previously proposed a device identification framework using device parameters. In this paper, we improve the identification accuracy of the identification using characteristic curves instead of device parameters. Through experiments using commercial quality SiC power MOSFETs, we demonstrate that 220 MOSFETs are successfully identified with an accuracy of 99.4, and the use of characteristic curves improved the device identification accuracy by 1.2 when compared to the conventional method.