A Novel Evaluation Metric For Deep Learning-Based Side Channel Analysis And Its Application

Since timing attack was proposed,side channel analysis(SCA)has shown great potential in breaking cryptosystems.By introducing side channel information to the analysis process,SCA can even outperform traditional cryptanalysis methods.Recently,deep learning techniques are widely used in SCA and show equivalent and even better performance compared to traditional methods.However,when and why deep learning techniques are effective and efficient for SCA are still unknown to researchers.Masure et al.illustrated that deep learning paradigm is suitable for evaluating implementations against SCA from a worst-case scenario point of view,yet their work is limited to balanced data and a specific loss function.Besides,there are critical conflicts between deep learning metrics and side channel metrics.In most cases,deep learning metrics are deceptive in foreseeing the feasibility and complexity of mounting a successful side channel attack,especially for imbalanced data.In the meantime,side channel metrics have very high complexity which makes it hard to be embedded in deep learning framework smoothly.To mitigate the gap between deep learning metrics and side channel metrics,we propose a novel Cross Entropy Ratio(CER)metric to evaluate the performance of deep learning models for SCA.CER is closely related to traditional side channel metrics Guessing Entropy(GE)and Success Rate(SR)and fits to deep learning scenario.Besides,we show that it works stably while deep learning metrics such as accuracy becomes rather unreliable when the training data tends to be imbalanced.However,estimating CER can be done as easy as natural metrics in deep learning algorithms with low computational complexity.Furthermore,we adapt CER metric to a new kind of loss function,namely CER loss function,designed specifically for deep learning in side channel scenario.In this way,we link directly the SCA objective to deep learning optimization.Our experiments on several datasets show that,for SCA with imbalanced data,CER loss function outperforms Cross Entropy loss function in various conditions.