When CPA and MIA go hand in hand

Correlation Power Analysis (CPA) is very fast and effective faced with linearity between the power consumption leakage and the power consumption model. Nevertheless its efficiency degrades quickly when the linearity between the leakage and the model is not verified for certain S-boxes. Mutual Information Analysis (MIA) is supposed to cope better with non-linearity. Still we encounter some difficulties when we estimate the mutual information. In this paper, we propose three approaches to enhance CPA and MIA for a given power consumption model (the Hamming weight model in our case). Two new attacking methods are suggested. We name them Extended Correlation Analysis (ECPA) and Combined Side-Channel Analysis (CSCA). They are not more complicated to practice than CPA or MIA. Experimental results show that they make considerable enhancements of CPA and MIA without much implementation or running time cost additions.