One of the important aspects of the security of block ciphers is the randomness of the cipher text. One criterion used to evaluate the Advanced Encryption Standard (AES) candidate algorithms was their demonstrated suitability as random number generators.
In this paper, we introduced a new approach to interpret the results of tests of randomness. This approach relies on using DieHard battery of tests which was designed originally to test the randomness of Random Number Generators (RNGs). The proposed approach was used to test the randomness of five types of data; plaintext, 3DES-encrypted, AES-encrypted, Serpent-encrypted, and Blowfish-encrypted. The tests resulted in 47 p-values for each type of data. This paper suggests the classification of results of DieHard tests into three areas; Safe Area, Doubt Area, and Failure Area. The resulting p-values for each data type were distributed over these areas according to the suggested ranges. The data type having more p-values in the Safe Area indicate better randomness, while the existence of many p-values in the Failure Area indicates deviation from randomness. The results of the implemented tests showed that AES- and Blowfish-encrypted data provided equal results in term of number of p-values distributed over different areas. The AES and Blowfish results were slightly better than Serpent-encrypted data while 3DES encrypted data had more p-values in the Doubt Area.
Mohammed M. Alani, Testing Randomness in Ciphertext of Block-Ciphers Using DieHard Tests, International Journal of Computer Science and Network Security (IJCSNS), Vol.10, No.4, April 2010, pp. 53-57.