Ranking Loss: Maximizing the Success Rate in Deep Learning Side-Channel Analysis

Abstract

The side-channel community recently investigated a new approach, based on deep learning, to significantly improve profiled attacks against embedded systems. Compared to template attacks, deep learning techniques can deal with protected implementations, such as masking or desynchronization, without substantial preprocessing. However, important issues are still open. One challenging problem is to adapt the methods classically used in the machine learning field (e.g. loss function, performance metrics) to the specific side-channel context in order to obtain optimal results. We propose a new loss function derived from the learning to rank approach that helps preventing approximation and estimation errors, induced by the classical cross-entropy loss. We theoretically demonstrate that this new function, called Ranking Loss (RkL), maximizes the success rate by minimizing the ranking error of the secret key in comparison with all other hypotheses. The resulting model converges towards the optimal distinguisher when considering the mutual information between the secret and the leakage. Consequently, the approximation error is prevented. Furthermore, the estimation error, induced by the cross-entropy, is reduced by up to 23%. When the ranking loss is used, the convergence towards the best solution is up to 23% faster than a model using the cross-entropy loss function. We validate our theoretical propositions on public datasets.

Breaking Masked Implementations of the Clyde-Cipher by Means of Side-Channel Analysis

“…Compared to the interpretation of deep learning based side-channel analysis as a classification problem that is more popular in the literature on deep learning based side-channel analysis (see e.g. [24,26] and the references cited therein), this has the advantage that the values we are trying to predict are in fact not categorical labels, but numbers. 5 By using a network that is a numerical regressor instead of a classifier, we can use that information to judge the severity of prediction errors, to alleviate the problem of rare output classes, and to induce an inductive bias towards mapping similar power traces to nearby Hamming weight estimates.…”

Towards Strengthening Deep Learning-based Side Channel Attacks with Mixup

“…They further adapted the CER metric to a new loss function to adopt in the training process, which can yield enhanced attack performance when dealing with imbalanced data. Another innovative loss function, Ranking Loss (RkL), was proposed by adapting the "Learning to Rank" approach in the Information Retrieval field to the side-channel context [17], which helps prevent approximation and estimation errors, induced by the typical cross-entropy loss.…”

“…MLPs joint the features in the dense layers and accordingly imitate the effect of higher-order attacks, hence standing as a natural choice when dealing with masking countermeasures [11], [12]. On the other hand, CNNs have the property of spatial invariance, bringing about intrinsic resilience against trace misalignment [13]- [17]. Furthermore, the internal architecture of CNNs make them adequate for automatical feature extraction from high-dimensional data, which eliminates the demand for preprocessing.…”

Focus is Key to Success: A Focal Loss Function for Deep Learning-Based Side-Channel Analysis

“…On the other hand, launching practical attacks and averaging the key rank to estimate the guessing entropy is computationally costly (especially if also done during the training phase, see, e.g., [14,19]). Consequently, the SCA community put a significant attention on developing SCA-specific metrics and loss functions, such as Cross-Entropy Ratio loss (CER) [30] and ranking loss (RKL) [28].…”