[Resource Topic] 2022/192: SoftSpokenOT: Communication--Computation Tradeoffs in OT Extension

Welcome to the resource topic for 2022/192

SoftSpokenOT: Communication–Computation Tradeoffs in OT Extension

Authors: Lawrence Roy


Given a small number of base oblivious transfers (OTs), how does one generate a large number of extended OTs as efficiently as possible? The answer has long been the seminal work of IKNP (Ishai et al., Crypto 2003) and the family of protocols it inspired, which only use Minicrypt assumptions. Recently, Boyle et al. (Crypto 2019) proposed the Silent-OT technique that improves on IKNP, but at the cost of a much stronger, non-Minicrypt assumption: the learning parity with noise (LPN) assumption. We present SoftSpokenOT, the first OT extension to improve on IKNP’s communication cost in the Minicrypt model. While IKNP requires security parameter \lambda bits of communication for each OT, SoftSpokenOT only needs \lambda / k bits, for any k, at the expense of requiring 2^{k-1} / k times the computation. For small values of k, this tradeoff is favorable since IKNP-style protocols are network-bound. We implemented SoftSpokenOT and found that our protocol gives almost a 5 \times speedup over IKNP in the LAN setting. Our technique is based on a novel silent protocol for vector oblivious linear evaluation (VOLE) over polynomial-sized fields. We created a framework to build maliciously secure 1-of-N OT extension from this VOLE, revisiting the existing work for each step. Along the way, we found several flaws in the existing work, including a practical attack against the consistency check of Patra et al. (NDSS 2017), while also making some improvements.

ePrint: https://eprint.iacr.org/2022/192

See all topics related to this paper.

Feel free to post resources that are related to this paper below.

Example resources include: implementations, explanation materials, talks, slides, links to previous discussions on other websites.

For more information, see the rules for Resource Topics .