InfoNEAT: Information Theory-based NeuroEvolution of Augmenting Topologies for Side-channel Analysis

The leakage of security-critical information has become an ever-increasing problem for cryptographic systems. Profiled side-channel analysis (SCA) leverages this leakage to extract sensitive data (e.g., the secret key) from cryptographic implementations. When combined with advanced methods in neural networks (NNs), profiled SCA can successfully attack even those crypto-cores assumed to be protected against SCA. Nevertheless and despite the rise in the number of studies devoted to NN-based SCA, a range of questions remain unanswered, namely: how to choose an NN with an adequate size, how to tune hyperparameters in the NN, when to stop the training, and how to explain the performance of the NN model in quantitative terms, in the context of SCA. Our proposed approach, "InfoNEAT," tackles these issues in a natural way. InfoNEAT relies on the concept of evolution of NNs (both the network architecture and parameters, so-called nueroevolution), enhanced by incorporating information-theoretic metrics to guide the evolution, halt it with a novel stopping criteria, and improve the time-complexity and memory footprint of it. Interestingly enough, our approach can be employed in various domains, although InfoNEAT, as presented in this paper, is tailored to the specific requirements of SCA. The performance of InfoNEAT is evaluated by applying that against publicly available datasets composed of real side-channel measurements. According to our assessment and on the basis of our results, InfoNEAT achieves the same performance (in terms of key recovery) as the state-of-the-art NN-based SCA. In addition to the considerable advantages regarding the automated configuration of NNs, InfoNEAT also demonstrates significant improvements over other approaches: the number of epochs and width of the NN (i.e., the number of nodes in a layer) is reduced at least by factor 1.25, and 6.66, respectively.