Paper 2014/583

Template Attacks Based On Priori Knowledge

Guangjun Fan and Yongbin Zhou and Dengguo Feng

Abstract

Template Attacks consist of two stages, the profiling stage and the extraction stage. In order to improve the classification performance of Template Attacks, a feasible and usual way is to characterize signals and noises more accurately. Under the assumption that a reference device is fully controlled by the attacker, in the profiling stage, the attacker can operate the reference device as many times as possible and samples a large number of actual power traces to help accurately characterize signals and noises at different interesting points. However, in some practical scenarios, this is not always the case and the attacker can only have access to a limited number of actual power traces. In this paper, we show that the attacker can still make Template Attacks practical and more powerful in the above scenario if he could obtain the priori knowledge about the reference device. The priori knowledge is just a kind of prior distribution of the actual value of the signal component of the instantaneous power consumption, rather than an accurate estimation of it. Evaluation results exhibit that leaking this kind of priori knowledge poses serious threat to the physical security of cryptographic devices.