Abstract: Camellia, as the final winner of 128-bit block cipher in NESSIE, is the most secure block cipher of the world. In 2003, Tsunoo proposed a Cache Attack using a timing of CPU cache, successfully recovered Camellia-128 key within 228 plaintexts and 35 minutes. In 2004, IKEDA YOSHITAKA made some further improvements on Tsunoo’s attacks, recovered Camellia-128 key within 221.4 plaintexts and 22 minutes. All of their attacks are belonged to timing driven Cache attacks, our research shows that, due to its frequent S-box lookup operations, Camellia is also quite vulnerable to access driven Cache timing attacks, and it is much more effective than timing driven Cache attacks. Firstly, we provide a general analysis model for symmetric ciphers using S-box based on access driven Cache timing attacks, point out that the F function of the Camellia can leak information about the result of encryption key XORed with expand-key, and the left circular rotating operation of the key schedule in Camellia has serious designing problem. Next, we present several attacks on Camellia-128/192/256 with and without FL/FL-1. Experiment results demonstrate: 500 random plaintexts are enough to recover full Camellia-128 key; 900 random plaintexts are enough to recover full Camellia-192/256 key; also, our attacks can be expanded to known ciphertext conditions by attacking the Camellia decryption procedure; besides, our attacks are quite easy to be expanded to remote scenarios, 3000 random plaintexts are enough to recover full encryption key of Camellia-128/192/256 in both local and campus networks. Finally, we discuss the reason why Camellia is weak in this type of attack, and provide some advices to cipher designers for hardening ciphers against cache timing attacks.
Category / Keywords: Camellia-128/192/256; block cipher; access driven; Cache timing attack; side channel attack; remote attack; F function; S-box lookup index; left circular rotating operation; key schedule; known ciphertext
第二篇：
“An Improved Differential Fault Attack on Camellia”
网页链接 http://eprint.iacr.org/2009/585
PDF链接 http://eprint.iacr.org/2009/585.pdf
Cryptology ePrint Archive: Report 2009/585
An Improved Differential Fault Attack on Camellia
ZHAO Xin-jie, WANG Tao
Abstract: The S-box lookup is one of the most important operations in cipher algorithm design, and also is the most effective part to prevent traditional linear and differential attacks, however, when the physical implementation of the algorithm is considered, it becomes the weakest part of cryptosystems. This paper studies an active fault based implementation attack on block ciphers with S-box. Firstly, it proposes the basic DFA model and then presents two DFA models for Feistel and SPN structure block ciphers. Secondly, based on the Feistel DFA model, it presents several improved attacks on Camellia encryption and proposes new attacks on Camellia key schedule. By injecting one byte random fault into the r-1th round left register or the the r-1th round key, after solving 8 equations to recover 5 or 6 propagated differential fault of the rth round left register, 5 or 6 bytes of the rth equivalent subkey can be recovered at one time. Simulation experiments demonstrate that about 16 faulty ciphertexts are enough to obtain Camellia-128 key, and about 32, 24 ciphertexts are required to obtain both Camellia-192/256 key with and without FL/FL-1 layer respectively. Compared with the previous study by ZHOU Yongbin et. al. by injecting one byte fault into the rth round left register to recover 1 equivalent subkey byte and obtaining Camellia-128 and Camellia-192/256 with 64 and 96 faulty ciphertexts respectively, our attacks not only extend the fault location, but also improve the fault injection efficiency and decrease the faulty ciphertexts number, besides, our DFA model on Camellia encryption can be easily extended to DFA on Camellia key schedule case, while ZHOU’s can not. The attack model proposed in this paper can be adapted into most of the block ciphers with S-boxes. Finally, the contradictions between traditional cryptography and implementation attacks are analyzed, the state of the art and future directions of the DFA on Block ciphers with S-boxes are discussed.
