Cryptology ePrint Archive: Report 2013/274

A time series approach for profiling attack

Liran Lerman and Gianluca Bontempi and Souhaib Ben Taieb and Olivier Markowitch

Abstract: The goal of a profiling attack is to challenge the security of a cryptographic device in the worst case scenario. Though template attack are reputed as the strongest power analysis attack, they effectiveness is strongly dependent on the validity of the Gaussian assumption. This led recently to the appearance of nonparametric approaches, often based on machine learning strategies. Though these approaches outperform template attack, they tend to neglect the time series nature of the power traces. In this paper, we propose an original multi-class profiling attack that takes into account the temporal dependence of power traces. The experimental study shows that the time series analysis approach is competitive and often better than static classification alternatives.

Category / Keywords: side-channel attack, power analysis, machine learning, time series classification.

Date: received 13 May 2013, last revised 13 May 2013, withdrawn 21 Feb 2017

Contact author: llerman at ulb ac be

Available format(s): (-- withdrawn --)

Version: 20170221:125700 (All versions of this report)

Short URL:

[ Cryptology ePrint archive ]