Efficient Dissimilarity Detection in Time Series With Application to Side-Channel Analysis

Abstract

This paper proposes a dissimilarity detection algorithm that can find different regions between similar signals. In particular, we address the detection problem of branching statements in program control flow using electromagnetic (EM) side-channels, where we have shown that such statements can be detected in emanating side-channels. Based on the findings, we have proposed a generalized approach for dissimilarity detection that can be efficiently applied to various real-world applications. In the proposed method, symbolic representation of the signals is used for efficient processing, where each signal frame is transformed into a string. The codebook of observed patterns is constructed with the reference signal. Then, the sequence of strings that is obtained from the main signal is compared with the codebook to find the newly observed patterns. Finally, the presented method outputs the samples of dissimilar regions in the signal compared to the reference. In the experiments, various EM side-channel signals are collected from different devices and different control flow examples to show the applicability and efficiency of the proposed method and the results show that dissimilarities can be detected with >98% accuracy.

Keywords

• Dissimilarity detection
• electromagnetic side-channel
• pattern matching
• program control flow
• time series representation