Design and Analysis of Neural Distinguisher for Differential Cryptanalysis of Lightweight Block Ciphers

Abstract

The design and analysis of small, low-power electronics, with limited memory and processing power is a niche area of research because of its wide range of applications. It is necessary to secure these devices and Lightweight ciphers will provide security for these devices (IoT, sensors, RFID). This paper deals with the neural distinguishers of lightweight block ciphers. A neural distinguisher scheme based on ResNet, Transformer, and a hybrid ResNet–Transformer model architectures for the ciphertexts is designed and the efficiency is demonstrated by considering the outputs of four lightweight block ciphers PRESENT, HIGHT, PRINCE, and TWINE. The results are obtained in the round-reduced environment and will provide empirical insights for the full-fledged differential cryptanalysis.