An Extended AES Scheme for Increasing Randomness in Ciphertext

Abstract: Technological advancements, such as high-speed internet, have transformed the world into a global village, raising concerns about privacy and secrecy amid cyberattacks and the disclosure of sensitive data. Cryptography and steganography are two well-known methods of secret communication. The former distorts the message, whilst the latter hides the very existence of the information within seemingly innocent carriers. Steganography faces challenges of steganalysis, whilst cryptography faces challenges of cryptanalysis. The extensive approval of Advanced Encryption Standard (AES) as an efficient symmetric cryptographic technique and other state- of-the-art data protection techniques has exposed them to increased attacks, prompting researchers to enhance AES’s strength. To contribute to the line of research, a novel matrix-based diffusion layer for the AES (MDLAES) scheme is proposed. The proposed scheme combines matrix data manipulation with the AES algorithm, adding an extra layer of security. This extended scheme produces a data scrambling algorithm that reconstructs plain text and secret keys before performing AES encryption on the result. The approach, first and foremost, ensures that knowledge of the initial key is insufficient to break the system; it also introduces a higher degree of randomness than the traditional AES cryptosystem. The study examined the performance of encryption and decryption operations using key sizes from 128 to 256 bits. As key size increases, CPU time and memory usage increase. It is also observed that AES encryption with matrix operations requires more CPU time and memory than the traditional AES algorithm. The research improves the diffusion rate by 3.04 when a single simulation is matched with the orthodox AES algorithm, and by 1.62 on average when 10 simulations are run with different keys. It is worth noting that a high diffusion rate and a double key make it more difficult for a plain-text attack.