Research Of Cache Attack On AES Encryption Algorithm

Along with the fast development of the international Internet, we are moving on the information superhighway. But the security issues during the communicating actives are growing at an astronomical rate, which attract increasing attention in recent years. Cache attack is one of side-channel attacks applied in cryptanalysis, it exploits the various features of Cache operation results, namely Cache-hit and Cache-miss. Because of the high attack efficiency and the simplicity, Cache attack has become indispensable in security information analysis.This paper proposes a side-channel attack model based on Cache attack, against symmetric encryption algorithms, such as AES(Advanced Encryption Standard). This model is brought forward in such research background that the same type of attack methods is invalid, and it is the first to quantify the attack efficiency mathematically. This attack model leverages the Padding Oracle weakness, which is associated with security defects in block cipher algorithms operating in CBC(Cipher-Block Chaining) mode. This flaw is one of CCA(Chosen Ciphertext Attack) security vulnerabilities because the cryptographic scheme using CBC is weak if an attacker has the ability to decrypt any ciphertext. The Cache attack module solves the key problems in current Padding Oracle attack: it may be impossible to obtain the oracle message which is crucial to the attack success. The attack model proposed in this paper can monitor the execution flow of the instructions related to generating the Oracle generation. Then useful information can be inferred by analyzing detailed algorithm process. At last, some or even all of plaintext can be recovered directlyAttack experiment targeting single byte of ciphertext to recover plaintext was conducted in two scenarios: a single machine and cross-VMs(Virtual Machine), aiming at cipher decrypting in Gnu TLS(a famous open source library). The result of the single machine experiment revealed that success rate for recover one bite could reach 99.5%,which is more efficient than the latest attack. Meanwhile cross-VMs experiment showed that plaintext could also be recovered with higher success rate by constructing complicated ciphertext chain. These scenarios experiment demonstrated the validity and high-efficiency of the attack model. In the last part, several countermeasures are suggested to resist attacks based on the attack model.