Device identification from mixture of measurable characteristics

In this paper, a novel framework that realizes identification of power MOSFETs for enhancing their security is proposed. Selecting mixture of measurable device parameters of MOSFETs as a feature vector, individual devices are robustly distinguished. In the experiments using a 11-parameter feature vector with machine learning techniques of random forest and support vector machine (SVM) as example classifiers, 70 commercial quality SiC power MOSFETs were successfully distinguished at 98.1% accuracy.