SIKE’d Up: Fast Hardware Architectures for Supersingular Isogeny Key Encapsulation
Brian Koziel, A-Bon Ackie, Rami El Khatib, Reza Azarderakhsh, Mehran Mozaffari Kermani
Abstract
Brian Koziel, A-Bon Ackie, Rami El Khatib, Reza Azarderakhsh, Mehran Mozaffari Kermani
Abstract
In this work, we present a fast parallel architecture to perform supersingular isogeny key encapsulation (SIKE). We propose and implement a fast isogeny accelerator architecture that uses fast and parallelized isogeny formulas. On top of our isogeny accelerator, we build a novel architecture for the SIKE primitive, which provides both quantum and IND-CCA security. We synthesized this architecture on the Xilinx Artix-7, Virtex-7, and Kintex UltraScale+ FPGA families. Over Virtex-7 FPGA's, our constant-time implementations are roughly 14% faster than the state-of-the-art with a better area-time product. At the NIST security level 5 on a Kintex UltraScale+ FPGA, we can execute the entire SIKE protocol in 15.3 ms. This work continues to improve the speed of isogeny-based computations and also features all parameter sets of the SIKE round 2 specification, with results applicable to NIST's post-quantum standardization process.
OpenAlex reports 78 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In this work, we present a fast parallel architecture to perform supersingular isogeny key encapsulation (SIKE). We propose and implement a fast isogeny accelerator architecture that uses fast and parallelized isogeny formulas. On top of our isogeny accelerator, we build a novel architecture for the SIKE primitive, which provides both quantum and IND-CCA security. We synthesized this architecture on the Xilinx Artix-7, Virtex-7, and Kintex UltraScale+ FPGA families. Over Virtex-7 FPGA's, our constant-time implementations are roughly 14% faster than the state-of-the-art with a better area-time product. At the NIST security level 5 on a Kintex UltraScale+ FPGA, we can execute the entire SIKE protocol in 15.3 ms. This work continues to improve the speed of isogeny-based computations and also features all parameter sets of the SIKE round 2 specification, with results applicable to NIST's post-quantum standardization process.
Key concepts: Isogeny, Key encapsulation, NIST, Computer science, Virtex, Post-quantum cryptography, Supersingular elliptic curve, Field-programmable gate array