Mutual information analysis for side channel attacks

Mutual information analysis is a recently proposed method in side channel analysis. In this thesis, a literature study of theoretical developments and current state of mutual information analysis is provided. Furthermore, two methods for mutual information estimation are assessed: histograms and cumulants. A Java-implementation of mutual information analysis is developed and design decisions are discussed. Results on the performance of mutual information analysis on both simulations and practical cryptographic embeddings, obtained using the implemented module, are provided, as well as a theoretical and practical comparison of mutual information analysis with correlation power analysis. Mutual information analysis is shown to be of specific interest on practical high security cryptographic embeddings where predictions and power consumption observations show nonlinear dependence.