EFFICIENT CNN-BASED PROFILED SIDE CHANNEL ATTACKS

Profiled side-channel attacks are now considered as powerful forms of used to break the security cryptographic devices. A recent line research has investigated a new profiled attack based on deep learning and many them have convolution neural network (CNN) architecture for attack. The effectiveness is greatly influenced by CNN architecture. However, current often been image recognition fields, choosing right architectures parameters adaption still challenging. In this paper, we propose an efficient unprotected masking-protected devices two architectures, called CNNn, CNNd respectively. Both proposed in paper property points interest power trace further determined Grey Wolf Optimization (GWO) algorithm. To verify attacks, experiments were performed set collected from Atmega8515 smart card when it performs AES-128 encryption, DPA contest v4 dataset ASCAD public