Paper 2018/196

Non-Profiled Deep Learning-Based Side-Channel Attacks

Benjamin Timon

Abstract

Deep Learning has recently been introduced as a new alternative to perform Side-Channel analysis. Until now, studies have been focused on applying Deep Learning techniques to perform Profiled Side-Channel attacks where an attacker has a full control of a profiling device and is able to collect a large amount of traces for different key values in order to characterize the device leakage prior to the attack. In this paper we introduce a new method to apply Deep Learning techniques in a Non-Profiled context, where an attacker can only collect a limited number of side-channel traces for a fixed unknown key value from a closed device. We show that by combining key guesses with observations of Deep Learning metrics, it is possible to recover information about the secret key. The main interest of this method, is that it is possible to use the power of Deep Learning and Neural Networks in a Non-Profiled scenario. We show that it is possible to exploit the translation-invariance property of Convolutional Neural Networks against de-synchronized traces and use Data Augmentation techniques also during Non-Profiled side-channel attacks. Additionally, the present work shows that in some conditions, this method can outperform classic Non-Profiled attacks as Correlation Power Analysis. We also highlight that it is possible to target masked implementations without leakages combination pre-preprocessing and with less assumptions than classic high-order attacks. To illustrate these properties, we present a series of experiments performed on simulated data and real traces collected from the ChipWhisperer board and from the ASCAD database. The results of our experiments demonstrate the interests of this new method and show that this attack can be performed in practice.