Non-Profiled Deep Learning-Based Side-Channel Analysis with Only One Network Training

We propose efficient protocols for non-profiled deep learning-based side-channel analysis (DL-SCA). While the existing protocol, proposed by Timon in 2019, requires computational resources training as many neural networks number of key candidates, our protocol only one network, which can be transformed into a network associated with each candidate. For instance, case AES, complexity is 1/256 that protocol. In this study, we describe idea and formulate it two depending on metrics used. numerically examine them implementing architectures, multilayer perceptron convolutional network. Using publicly available open data (ASCAD), show both efficiently work expected. also clarify trained Timon’s original case, recycled an attack against same device different materials. Non-profiled DL-SCAs are superior to profiled ones they require no reference profiling before analyzing target device. This property holds proposal well.