Moments-Correlating DPA
Amir Moradi, François‐Xavier Standaert · 2016
We generalize correlation-enhanced power analysis collision attacks into moments-correlating DPA. The resulting distinguisher is applicable to the profiled and non-profiled (collision) settings and is able to exploit information lying in any statistical moment. It also benefits from a simple rule-of-thumb to estimate its data complexity. Experimental results show that such a tool allows answering with confidence to some important questions regarding the design of side-channel countermeasures (e.g. what is the most informative statistical moment in the leakages of a threshold implementation). We further argue that moments-correlating DPA is a natural candidate for leakage detection tests, enjoying the simplicity of correlation power analysis and advanced features for the evaluation of higher-order attacks with an easy-to-compute confidence level.