Deep-Learning based Nonprofiling Side-Channel Attack on Mask leakage-free Environments using Broadcast Operation

With the recent development of artificial intelligence (AI), efforts to apply related technologies various fields are rapidly increasing. In field cryptanalysis, research utilizing deep learning is continuously being published in order keep up with this trend. Side-channel analysis a type cryptanalysis that uses physical information and can be classified into profiling nonprofiling analyses. Nonprofiling attacks using take advantage fact training performed relatively well when right key compared wrong key. Masking countermeasures applied design secure cipher against side-channel analysis. The traditional second-order attack for analyzing masked ciphers used by preprocessing side channel remove mask value. However, has able omit process. Related works proposed so far attempted analyze cipher, but focused only on 1-byte masking itself. reality, grasping time-points, which revealed, difficult from secret area. study, we attempt case combining 2-byte information, not information. We also propose neural network scheme perform more effective attacks. method highlights relative difference between keys. Previous evaluation criteria been lacking. Therefore, herein new metrics easily demonstrate their validity simulation actually collected data. As result experiment, methods based loss metric improved approximately 228.59% dataset 739.46% real binary labeling. And it reduced minimum number analytical traces 78.95% 72.5%, respectively.