Artificial Intelligence Applied to the Identification of Block Ciphers under CBC Mode

Abstract

This research introduces a novel methodology for identifying symmetric cryptosystems operating in Cipher Block Chaining (CBC) mode based solely on encrypted texts. The approach combines statistical tests from NIST STS with machine learning algorithms, analyzing DES, 3DES, Blowfish, Camellia, and AES. The experimental results demonstrate an 84% identification rate for multiclass identification using random keys and initialization vectors. These findings are valuable in the field of information security and aid in minimizing cryptanalytic efforts.