TESLAC: Accelerating Lattice-Based Cryptography with AI Accelerator

• Published in: Security and Privacy in Communication Networks Vol. 398; pp. 249 - 269
• Main Authors: Wan, Lipeng, Zheng, Fangyu, Lin, Jingqiang
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2021; Springer International Publishing
• Series: Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering
• Subjects: AI accelerator; LAC; Lattice-based cryptosystems; Polynomial multiplication over rings; Tensor Core
• ISBN: 9783030900182; 3030900185
• ISSN: 1867-8211; 1867-822X
• DOI: 10.1007/978-3-030-90019-9_13

In this paper, we exploit AI accelerator to implement cryptographic algorithms. To the best of our knowledge, it is the first attempt to implement quantum-safe Lattice-Based Cryptography (LBC) with AI accelerator. However, AI accelerators are designed for machine learning workloads (e.g., convolution operation), and cannot directly deliver their strong power into the cryptographic computation. Noting that polynomial multiplication over rings is a kind of time-consuming computation in LBC, we utilize a straightforward approach to make the AI accelerator fit well for polynomial multiplication over rings. Additional non-trivial optimizations are also made to minimize the overhead of transformation, such as using low-latency shared memory, coalescing memory access. Moreover, based on NVIDIA AI accelerator, Tensor Core, we have implemented a prototype system named TESLAC and give a set of comprehensive experiments to evaluate its performance. The experimental results show TESLAC can reach tens of millions of operations per second, achieving a performance speedup of two orders of magnitude from the AVX2-accelerated reference implementation. Particularly, with some techniques, TESLAC can also be scaled to other LBC with larger modulo q.