Time, Memory and Accuracy Tradeoffs in Side-Channel Trace Profiling.

Template attacks are one of the most powerful classes of side-channel attacks. Template attacks begin with an offline step, in which the side-channel traces are profiled, and decoders are created for each side-channel leak. In this paper, we analyze the compression step of the trace profiling process. This compression step, which is a central part of the decoder’s training process, is used to reduce the amount of time, memory consumption, and data required to successfully perform the attack; various practical methods have been proposed for this step, including one which uses an efficient means both for selecting the points of interest (POI) in the power trace and for preprocessing noisy data. We investigate ways to improve the efficiency of the attack by implementing several compression methods which select the most informative power consumption samples from power traces. We develop a unique dedicated evaluation system to compare the performance of various decoders with different compression methods on real-world power traces. Our findings indicate that our proposed decoder for side-channel traces outperforms the current state of art in terms of speed, resource consumption, and accuracy. We also demonstrate our decoder’s effectiveness under resource-constrained conditions, and show that it achieves over 70% accuracy even if there are fewer than 1,000 traces in the profiling phase.