Neural Cryptanalysis of Lightweight Block Ciphers Using Residual MLPs

The security of Internet of Things (IoT) devices is a growing concern, given their widespread deployment in environments with limited computational and energy resources. Lightweight block ciphers, such as SIMON and SPECK, are designed to provide efficient cryptographic operations while minimizing computational overhead. However, evaluating their resilience against emerging attack vectors is vital for maintaining robust protection. This paper introduces a neural cryptanalysis approach for evaluating the security of SIMON and SPECK block ciphers, by leveraging a Residual Multi-Layer Perceptron (ResMLP) model in order to approximate the encryption and decryption processes. Experimental results demonstrate the effectiveness of the approach in revealing vulnerabilities, showcasing its efficiency and scalability in performing neural cryptanalysis on lightweight block ciphers.