Modeling attacks against device authentication using CMOS image sensor PUF

Abstract

A CMOS image sensor physical unclonable functions (CIS PUF) which generates unique response extracted from manufacturing process variation is utilized for device authentication. In this paper, we report modeling attacks to the CIS PUF, in which column fixed pattern noise is exploited in a sorting attack. When the PUF response is generated with pairwise comparison method, unknown responses are predicted with probability over 87.8% with only 0.31% training sample of whole challenge and response pairs.