Parallel and High Speed Hashing in GPU for Telemedicine Applications


Lee, Wai Kong and Phan, Raphael Chung Wei and Goi, Bok Min and Chen, Lanxiang and Zhang, Xiujun and Xiong, Naixue N. (2018) Parallel and High Speed Hashing in GPU for Telemedicine Applications. IEEE Access, 6. pp. 37991-38002. ISSN 2169-3536

[img] Text
Restricted to Repository staff only

Download (8MB)


With the advent of a telemedicine technology, many medical services can be provided remotely, which greatly enhances the welfare of our mankind. However, security and privacy of medical data transmitted through telecommunication systems remain a serious issue to be resolved when deploying such services. In particular, the medical images and data are stored in the cloud or transmitted over an insecure channel, may suffer from unauthorized modifications by malicious attackers. Hence, integrity of such medical data is of utmost importance for the telemedicine applications. Cryptographic hash functions (e.g., SHA-3) can be used to ensure the integrity of medical data communicated over the insecure channel. However, when the volume and size of medical data grow (e.g., high resolution medical image), it is difficult for conventional CPU-based system to hash these data in timely manner. In view of that, we are motivated to research on improved implementation techniques of the Keccak hash function in massively parallel platforms, as the result of such work can be used in improving the speed performance of the telemedicine applications. Graphical processing unit (GPU) is one of the emerging platforms with massively parallel processing power that can be harnessed to solve computational problems much faster than conventional CPUs. In this paper, we present the efficient implementation of tree-mode Keccak-f(1600) in GPU and investigate the effect of parallel granularities by hashing one copy of Keccak permutation function using 1 thread, 5 threads, and 25 threads, respectively. We also proposed a new technique to implement the treemode Keccak-f(1600) based on dynamic parallelism offered in new NVIDIA GPU. Our experimental results show that the parallel granularity of one thread produces the highest hash throughput at 28.51 Gb/s.

Item Type: Article
Uncontrolled Keywords: Graphics processing units, Biomedical imaging, Security, telemedicine, GPU, SHA-3
Subjects: T Technology > T Technology (General)
Divisions: Faculty of Engineering (FOE)
Depositing User: Ms Rosnani Abd Wahab
Date Deposited: 30 Nov 2020 08:51
Last Modified: 30 Nov 2020 08:51


Downloads per month over past year

View ItemEdit (login required)