A Quantitative Analysis of Non-Profiled Side-Channel Attacks Based on Attention Mechanism

In recent years, the deep learning method has emerged as a mainstream approach to non-profiled side-channel attacks. However, most existing methods of learning-based attack rely on traditional metrics such loss and accuracy, which often suffer from unclear results in practical scenarios. Furthermore, previous studies have not fully considered properties power traces long time-series data. this paper, novel architecture is proposed, incorporates attention mechanism derives corresponding metric. By attaching after network layers, provides quantitative prediction correct key. Moreover, can effectively extract analyze features traces. The success rate different datasets at least 86%, demonstrates superior reliability compared other works when facing various countermeasures noise. Notably, even scenarios where accuracy fail provide reliable results, proposed metric remains capable accurately distinguishing