Machine-Learning assisted Side-Channel Attacks on RNS-based Elliptic Curve Implementations using Hybrid Feature Engineering

Side-channel attacks based on machine learning have recently been introduced to recover the secret information from software and hardware implementations of mathematically secure algorithms. Convolutional Neural Networks (CNNs) have proven to outperform the template attacks due to their ability of handling misalignment in the symmetric algorithms leakage data traces. However, one of the limitations of deep learning algorithms is the requirement of huge datasets for model training. For evaluation scenarios, where limited leakage trace instances are available, simple machine learning with the selection of proper feature engineering, data splitting, and validation techniques, can be more effective. Moreover, limited analysis exists for public-key algorithms, especially on non-traditional implementations like those using Residue Number System (RNS). Template attacks are successful on RNS-based Elliptic Curve Cryptography (ECC), only if the aligned portion is used in templates. In this study, we present a systematic methodology for the evaluation of ECC cryptosystems with and without countermeasures against machine learning side-channel attacks using two attack models. RNS-based ECC datasets have been evaluated using four machine learning classifiers and comparison is provided with existing state-of-the-art template attacks. Moreover, we analyze the impact of raw features and advanced hybrid feature engineering techniques, along with the effect of splitting ratio. We discuss the metrics and procedures that can be used for accurate classification on the imbalance datasets. The experimental results demonstrate that, for ECC RNS datasets, the efficiency of simple machine learning algorithms is better than the complex deep learning techniques when such datasets are not so huge