Fast Homomorphic Encryption Based on CPU-4GPUs Hybrid System in Cloud

摘要

Security is an ever-present consideration for applications and data in the cloud computing environment. As an important method of performing computations directly on encrypted data without any need of decryption and compromising privacy, homomorphic encryption is an increasingly popular topic of protecting the privacy of data in cloud security research. However, as high computational complexity of the homomorphic encryption, it will be a heavy workload for computing resources in the cloud computing paradigm. Motivated by this observation, this paper proposes a fast parallel scheme with DGHV algorithm based on CPU-4GPUs hybrid system. Our main contribution of this paper is to present a parallel processing stream scheme for large-scale data encryption based on CPU-4GPUs hybrid system as fast as possible. Particularly, the proposed method applies CPU-4GPUs parallel implementation to accelerating encryption operation with DGHV algorithm to reduce the time duration and provide a comparative performance study. We also make further efforts to design a pipeline architecture of processing stream in CPU-4GPUs hybrid system to accelerate encryption for DGHV algorithm. The experiment results show that our method gains more than 91 % improvement (run time) and 70% improvement compared to the serial addition and multiplication operation with DGHV algorithm respectively.