Based On Multi-core Platforms Elliptic Curve Cryptographic Algorithm And The Aes Algorithm Parallelization

With the rapid development of informatization and internet, information security becomes more and more important. Symmetric cryptography and asymmetric cryptography are an important way to achieve and ensure information security. Symmetric cryptography has advantages of fast encryption and short key, but its disadvantage is difficult key distribution and management. Asymmetric cryptography can overcome that shortcoming, but it also has weakness of large amount of mathematical operations and slow encryption. The Advanced Encryption Standard (AES) based on block cipher and the Elliptic Curve Cryptography (ECC) based on discrete logarithm problem (DLP) are two well-known cryptographic systems, and they are very important for the research and development of cryptography. So a parallel base-point judgment algorithm for ECC over GF(p) and a parallel AES crypto algorithm based on different decomposition are investigated in order to improve the velocity of encryption and decryption.In this thesis, the analysis of existing parallel base-point judgment algorithm for ECC and the implementation for serial AES crypto algorithms are discussed, and a new parallel base-point judgment algorithm for ECC and parallel AES crypto algorithm based on different decomposition are proposed. All of the parallelization are based on multi-core technology and OpenMP multithreading programming technology.For the proposed parallel base-point judgment algorithm for ECC, we use the NAF form of scalar k, and utilize one core to do ECDBL(elliptic curve doubling) operation while another core to do ECADD(elliptic curve addition) operation. These two cores communicate to each other by a shared circular cache. The parallel efficiency is promoted significantly with a maximum of approximate to 110 %. Moreover, the experimental data between our algorithm and the related algorithm are compared. The results show that the speed of the base-point judgment algorithm is obviously improved. With the binary size of scalar increasing, the parallel efficiency decreases, and keeps stably in the end. In addition, the capacity of circular cache has no influence on parallel efficiency, and the algorithm offers better security against SCA. For the parallel AES crypto algorithm, two methods are used to parallel. The first method is based on task decomposition that the round functions are parallelized. The experiment results show that the time consuming is much longer than that of before. So that method is not feasible. That is because the time consuming of threads maintenance for OpenMP is much greater than the round operation. The second method is based on data decomposition. In this way, the Electronic Code Book is used which divides the plaintext into two equal part, and then the two cores compute the two part independently. It is shown that the algorithm has a nearly linear speed-up and the parallel efficiency has been steady at 100%, and the length of key and block has no effect on the parallel efficiency. So the algorithm based on data decomposition is feasible for parallel.