Chinese Journal of Computers   Full Text
Title	A New Statistical Test on Block Ciphers
Authors	CHEN Hua1) FENG Deng-Guo1) FAN Li-Min1)，2)
Address	1)(State Key Laboratory of Information Security， Institute of Software, Chinese Academy of Sciences, Beijing 100190) 2)(Graduate University of Chinese Academy of Sciences, Beijing 100039)
Year	2009
Issue	No.4(595—601)
Abstract &   Background	Abstract Statistical tests are playing very important part in the security evaluation on block ciphers, which have been applied by many cryptographic standard organizations in the evaluation process. In this paper, an effective and practical statistical test is proposed，which takes block size as the statistics unit. In the test, one state byte of a block is taken all possible values while other bytes are fixed, and after cryptographic transformation 256 output values are applied XOR(exclusive-or) operation. By testing whether every output bit of XORed value is 0 with probability 1/2, the randomness of block ciphers can be judged. The method can also somewhat reflect the resistance ability of a block cipher to integral attack. Meanwhile, based on the generalized integral attack, a more generalized statistical test is proposed on the basis of the given method. Moreover, the proposed statistical test is applied on Rijndael, Camellia and SMS4 algorithm, and good statistical property begins to behave respectively after 4 round, 5 round and 7 round operation. Keywords statistical test; block cipher; security evaluation; integral attack Background Cryptographic algorithms are playing core roles in information security,which can provide confidentiality, integrity and authenticity of information. During the design and analysis of cryptographic algorithms, how to evaluate the algorithms is becoming a key problem. There have existed many avenues to evaluate cryptographic algorithms, among which the statistical test is very important. Compared with cryptographic analysis, statistical tests have many advantages including high automation, generic test models and so on. Currently, many cryptographic standard plans have applied statistical tests in the process of the evaluation on cryptographic algorithms. The standard plans contain AES(Advanced Encryption Standard), NESSIE(New European Schemes for Signature, Integrity, and Encryption), ECRYPT(European Network of Excellence for Cryptology) and so on. To sum up, there are two kinds of statistical tests on cryptographic algorithms. The first one is the pure randomness test on the output sequences of cryptographic algorithms, under which cryptographic algorithms are thought of pseudorandom number generators. Up to now, there have existed a lot of randomness test methods such as frequence test, run test, poker test, autocorrelate test and so on. Besides the randomness test, another kind of statistical tests is based on the size and structure of cryptographic algorithms. Compared with the first one, the second one can detect the statistical weaknesses more easily. Unfortunately, the research on the second kind of statistical tests is still weak. In this paper, an effective and practical statistical test on block ciphers is proposed，which belongs to the second kind of statistical tests. In the test, one state byte of a block is taken all possible values while other bytes are fixed, and after cryptographic transformation 256 output values are applied XOR(exclusive-or) operation. By testing whether every output bit of XORed value is 0 with probability 1/2, the randomness of block ciphers can be judged. The method can also somewhat reflect the resistance ability of a block cipher to integral attack. The research work in this paper can be used in the future practical cryptographic evaluation on block ciphers.