Diffusion analysis of EFN-MDS structure

Abstract

In general, block ciphers consist of one top-level structural model into which the round function F is plugged into. The study focuses on Extended-Feistel-Network (EFN) that is a generalization of a Feistel Network (FN). This structure is employed in several ciphers that were developed for Advanced Encryption Standard such as CAST-256, MARS and RC6. The problem with EFN is that it requires many rounds when the number of sub-blocks used in EFN is large. This paper proposed a new structural model that can overcome this problem by incorporating EFN with a linear transformation based on Maximum Distance Separable (MDS) codes. The diffusion analysis shows that EFN-MDS requires at most half the number of rounds to achieve completeness property as compared to EFN structure. Therefore the proposed structure is suitable for designing ciphers with scalable block sizes and ciphers with large block sizes.