Exploring Feature Selection Scenarios for Deep Learning-based Side-Channel Analysis

One of the main promoted advantages of deep learning in profiling sidechannel analysis is the possibility of skipping the feature engineering process. Despite that, most recent publications consider feature selection as the attacked interval from the side-channel measurements is pre-selected. This is similar to the worst-case security assumptions in security evaluations when the random secret shares (e.g., mask shares) are known during the profiling phase: an evaluator can identify points of interest locations and efficiently trim the trace interval. To broadly understand how feature selection impacts the performance of deep learning-based profiling attacks, this paper investigates four different feature selection scenarios that could be realistically used in practical security evaluations. The scenarios range from the minimum possible number of features to the whole available trace samples. Our results emphasize that deep neural networks as profiling models show successful key recovery independently of explored feature selection scenarios against first-order masked software implementations of AES 128. Concerning the number of features, we found three main observations: 1) scenarios with less carefully selected point-ofinterest and larger attacked trace intervals are the ones with better attack performance in terms of the required number of traces during the attack phase; 2) optimizing and reducing the number of features does not necessarily improve the chances to find good models from the hyperparameter search; and 3) in all explored feature selection scenarios, the random hyperparameter search always indicate a successful model with a single hidden layer for MLPs and two hidden layers for CNNs, which questions the reason for using complex models for the considered datasets. Our results demonstrate the key recovery with a single attack trace for all datasets for at least one of the feature selection scenarios.