Learning-based Side-Channel Analysis on PIPO

Abstract

As the global IoT market increases, the importance of security in the IoT environment is growing. So, studies on lightweight cipher techniques are actively underway for limited environments. In ICISC 2020, PIPO, a bitslice lightweight cipher that can effectively apply a countermeasure considering side-channel analysis, was proposed. In this paper, we propose Deep Learning-based profiled and non-profiled Side-Channel Analysis for PIPO. In profiled attack, we use an 8-bit model instead of 1-bit model that considered the bitslice characteristic of S-Box output. Although an each bit of S-Box output is distributed across the power trace, the 8-bit model has shown high training performance with 98% accuracy, and was able to derive right key successfully. In non-profiled attack, we propose a labeling technique suitable for the bitslice characteristic and show the excellence of our proposed labeling through experiments. Also, we expect that these characteristics will apply to other bitslice block ciphers as well as PIPO.

This work was supported by Institute for Information & communications Technology Promotion (IITP) grant funded by the Korea government (MSIT) (No. 2017-0-00520, Development of SCR-Friendly Symmetric Key Cryptosystem and Its Application Modes).