Research On High-Speed Implementation Methods Of Block Ciphers Via GPU And Bitsliced

Block cipher is widely used to encrypt data to protect information security and privacy due to its convenience in software and hardware.How to implement high-speed block ciphers to provide security support in the current environment of big data and 5G communication with massive data has become a hot topic of discussion in academia and industry.The fast implementation of block ciphers can be optimized in both software and hardware aspects.The optimization in hardware is mainly realized by using graphics processing unit(GPU),and the algorithm is optimized based on the advantages of GPU's multi-threading and suitable for parallel computing;the optimization in software is mainly the application of bitsliced technology.Based on the idea of bitsliced technology to simulate the implementation of hardware in software implementation,the goal of improving implementation performance is achieved.Based on GPU and bitsliced technology,this thesis studies the high-speed implementation method of block cipher algorithm.The work content includes:1.A fast implementation method for SKINNY encryption algorithm under GPU is proposed.Based on the characteristics of the SKINNY cryptographic algorithm structure,the algorithm itself is optimized in parallel,and then the variable memory of the SKINNY algorithm is allocated for parallel implementation.The research results show that the efficiency and throughput of the SKINNY algorithm implemented by CUDA have been greatly improved.The ECB mode achieves maximum efficiency improvement of 99.85%and maximum speedup ratio of 671.The CTR mode achieves maximum efficiency improvement of 99.87% and maximum speedup ratio of 765.In particular,the throughput of the proposed SKINNY parallel algorithm has 1.29 times and 2.55 times of those of the existing AES and SKINNY parallel algorithms,respectively.2.A GPU parallel scheme for the national cryptographic algorithm SM4 is proposed.Through the analysis and implementation of the block cipher SKINNY algorithm,it is extended to the study of the national cryptographic algorithm SM4.Based on the structure and characteristics of the SM4 algorithm,combined with the characteristics of CUDA,through the introduction of page-locked memory and CUDA streams,the performance of SM4 is optimized and improved.The research results show that the SM4 algorithm finally obtains a maximum speedup of 89 and a maximum throughput of 31.41 Gbps.And compared with the existing SM4 parallel algorithm,the result of this thesis has a performance improvement of 3.44 times.3.A fast implementation scheme of AES algorithm based on fixsliced technology and GPU is proposed.CUDA and fixsliced technology are applied to the AES algorithm,and the linear layer is adjusted accordingly by fixing some slices,so as to improve the performance of the algorithm.Furthermore,full-fixsliced and semi-fixsliced are used to implement AES,and it is found that the time and data throughput of the two methods are not much different,but the semi-fixsliced implementation uses fewer operands.The research results show that:the application of fixsliced technology improves the performance of the AES algorithm and increase its space utilization;the AES algorithm is implemented using fixsliced technology,which has a performance improvement of 2.60 and 5.88 times under the CPU and GPU respectively,and the maximum speedup ratio reaches 310 under the GPU.And compared with the existing AES parallel algorithm,the result of this thesis has a performance improvement of up to 21.6 times.