Detecting similar code segments through side channel leakage in microcontrollers

Abstract

We present new methods for detecting plagiarized code segments using side-channel leakage of microcontrollers. Our approach uses the dependency of side-channel leakage on processed data and requires that the implementation under test accepts varying known input data. Detection tools are built upon a similarity matrix that contains the absolute correlation coefficient for each combination of time samples of the two possibly different implementations as result of side channel measurements. These methods are evaluated on smartcards with ATMega163 microcontroller using different test applications written in assembly language. We show that our methods are highly robust even against a skilled adversary who modifies the original assembly code in various ways. Our approach is non-intrusive, so that the application does not need to be additionally watermarked in order to be protected—the resulting pattern of data leakage of the microcontroller executing the code is considered as its own watermark.