Improving DPA by peak distribution analysis

Abstract

Differential Power Analysis (DPA) attacks extract secret key information from cryptographic devices by comparing power consumption with predicted values based on key candidates and looking for peaks which indicate a correct prediction. A general obstacle in the use of DPA is the occurrence of so called ghost peaks, which may appear when evaluating incorrect key candidates. Some ghost peaks can be expected from the structure and may actually leak information. We introduce a DPA enhancement technique-Euclidean Differential Power Analysis (EDPA), which makes use of the information leaked by the ghost peaks to diminish the ghost peaks themselves and bring forward the correct key candidate. The EDPA can be combined with any standard DPA attack irrespective of the distinguisher used. We illustrate that EDPA improves on DPA with both simulations and experiments on smart cards. © 2011 Springer-Verlag Berlin Heidelberg.