Paper 2021/1091

No (Good) Loss no Gain: Systematic Evaluation of Loss functions in Deep Learning-based Side-channel Analysis

Maikel Kerkhof and Lichao Wu and Guilherme Perin and Stjepan Picek

Abstract

Deep learning is a powerful direction for profiling side-channel analysis as it can break targets protected with countermeasures even with a relatively small number of attack traces. Still, it is necessary to conduct hyperparameter tuning for strong attack performance, which can be far from trivial. Besides a plethora of options stemming from the machine learning domain, recent years also brought neural network elements specially designed for side-channel analysis. An important hyperparameter is the loss function, which calculates the error or loss between the actual and desired output. The resulting loss is used to update the weights associated with the connections between the neurons or filters of the deep learning neural network. Unfortunately, despite being a highly relevant hyperparameter, there are no systematic comparisons among different loss functions. This work provides a detailed study on the performance of different loss functions in the SCA context. We evaluate five loss functions commonly used in machine learning and two loss functions proposed for SCA. Our results show that one of the SCA-specific loss functions (called CER) performs very well and outperforms other loss functions in most evaluated settings. Finally, our results show that categorical cross-entropy represents a good option for most settings, especially if there is a requirement to work well with different neural network architectures.