Security Analysis and Modeling Attacks on Duty Cycle Multiplexer PUF

Abstract

The security of strong physical unclonable function (PUF) such as arbiter PUF is becoming a question after machine learning method like logistic regression (LR) is proved to work well on modeling arbiter PUF. Duty cycle multiplexer (DC MUX) PUF is a newly proposed PUF design to enhance the reliability of arbiter PUF. The security of DC MUX PUF is analyzed in this paper, and the relation between error rate and the number of challenge-response pairs (CRPs) is obtained with the previous theoretical result. Then LR attack on DC MUX PUF is conducted, the results match the theoretical result. The prediction error of DC MUX PUF is both predicted and then proved to be higher than that of arbiter PUF. Besides, the training time of DC MUX PUF is 40000 times higher than arbiter PUF. As a result, DC MUX PUF is proved to be safer than arbiter PUF.