Sunday, April 20, 2014

Correlation Power Analysis (CPA)


Correlation Power Analysis Attacks
Correlation Power Analysis (CPA) is one of the most fable power based side channel analysis attacks. CPA attack was proposed by Brier et. al [1] in 2004 and works well even with noise. Unlike (Differential Power Analysis) DPA, CPA has a high convergence towards the correct hypothesis. CPA uses Pearson correlation coefficient algorithm to correlate the hypothetical power consumption with the data being processed. I have some figures to show in detail explanation of CPA but unpublished work cannot be included to my blog (Oh hell yeah. I am lazy to redraw figures for my papers again).
Pearson Correlation Coefficients can be varied from -1 to +1. Sometimes in the CPA graphs, the secret key behaves as a gradual descend and then a rise. This is because, we are not sure about the polarity of the correlation coefficients and sometimes the correlation coefficients are started from one end (+ve or -ve) and will converge to the other end. The absolute values of the correlation coefficients are plotted and that is the reason for descend-ascend behavior which makes a huge puzzle for researchers who use DPA.  I will post a CPA graph once the research is published.
[1]. Eric Brier, Christophe Clavier, Francis Olivier: Correlation Power Analysis with a Leakage Model. CHES 2004: 16-29

No comments:

Post a Comment