Reliable Information Extraction for Single Trace Attacks

Abstract

Abstract-Side-channel attacks using only a single trace crucially rely on the capability of reliably extracting side-channel information (e.g. Hamming weights of intermediate target values) from traces. In particular, in original versions of simple power analysis (SPA) or algebraic side channel attacks (ASCA) it was assumed that an adversary can correctly extract the Hamming weight values for all the intermediates used in an attack. Recent developments in error tolerant SPA style attacks relax this unrealistic requirement on the information extraction and bring renewed interest to the topic of template building or training suitable machine learning classifiers.In this work we ask which classifiers or methods, if any, are most likely to return the true Hamming weight among their first (say s) ranked outputs. We experiment on two data sets with different leakage characteristics. Our experiments show that the most suitable classifiers to reach the required performance for pragmatic SPA attacks are Gaussian templates, Support Vector Machines and Random Forests, across the two data sets that we considered. We found no configuration that was able to satisfy the requirements of an error tolerant ASCA in case of complex leakage.

Template Attacks vs. Machine Learning Revisited (and the Curse of Dimensionality in Side-Channel Analysis)

“…In our simulated setting, the most interesting gain is exhibited for RFbased models, thanks to their random feature selection. Interestingly, the recent work of Banciu et al reached a similar conclusion in a different context, namely, Simple Power Analysis and Algebraic Side-Channel Analysis [1]. Besides, and admittedly, the simulated setting we investigated is probably most favorable to TA, since only estimation errors can decrease the accuracy of the adversary/evaluator models in this case.…”

“…In this paper, we aim to complement these previous works with a more systematic investigation of the conditions under which ML-based attacks may outperform TA (or not) 1 . For this purpose, we start with the general intuition that ML-based approaches are generally useful in order to deal with high-dimensional data spaces.…”

Efficient Profiled Attacks on Masking Schemes

“…As a result, although we do not claim that RF is the universally best machine learning model for such contexts (as formalised by the no-free-lunch theorem), several works already report the high success probability of RF to extract the secret key from leakages while not requiring a large quantity of measurements neither a high time complexity for the selection of its structure [1], [16], [17], [20]. In this paper, we confirm these advantages on devices protected with a masking scheme.…”

Scramble Suit: A Profile Differentiation Countermeasure to Prevent Template Attacks

“…To this end, [22], [27] report that countermeasures effective against non-profiled side channel attacks are not sufficient to ward off against profiled ones. This has lead to practical attacks against real world devices [28], despite some may have very tight limitations on the number of collectible measurements [29]. A fundamental assumption of profiled attacks is that the side channel behavior derived on an instance of a device provides a good, ideally perfect, fit for the behavior of another instance of the same device.…”