SING: Improving the Efficiency of MPC Protocol Assignment using Graph Neural Networks

This repository contains the full codebase and dataset of SING: Improving the Efficiency of MPC Protocol Assignment using Graph Neural Networks, to appear at USENIX Security'26. The paper is available on ePrint: https://eprint.iacr.org/2026/1237.

The files included in this repository are:

• SING-main.tar.gz: This archive contains the complete source directory tree of our codebase, including vendored versions of the ABY and Silph dependencies. A README.md with installation and reproduction instructions is included in the archive root.
• dataset-full-c.tar.gz: This archive contains the full unprocessed dataset of MPC circuits as C source files
• dataset-full.tar.gz: This archive contains the full processed dataset of MPC circuits
• CHANGELOG.md: We will describe updates to this repository in this file where applicable