Cost-Efficient Parallel RSA Decryption with Integrated GPGPU and OpenCL

Secure communications become more, more important with the exponential growth of Internet data services. However, to massively deploy such services in current networks, we need to deal with different architectures, time-consuming encryption/decryption algorithms. Recently, numerous research efforts have been focused on parallel computing using high-end discrete GPGPU to achieve peak throughput, while the study on more cost-efficient integrated GPGPU is still open, which is included in almost all of the Intel CPUs currently, e.g. Celeron, Pentium, Core, even Xeon-E3 families. In this paper, we propose a generic heterogeneous computing approach for the decryption of RSA algorithm. Based on OpenCL, we are able to implement our idea on the off-the-shelf mini PC with low-cost integrated GPGPU. According to the experimental results, we show that the proposed scheme can fully utilize the capacity of integrated GPGPU,, a 2 times throughput improvement over the CPUonly solution can be expected.