CCS2026

Relect: Single Secret Leader Election via FHE with Reduced Computation and Communication and Transparent Setup

Haofei Liang, Zeyu Liu, Yunhao Wang, Xiang Xie, Yu Yu, Fan Zhang

摘要

Relect — Artifact for ACM CCS 2026 (#923) Paper: #923 Relect: Single Secret Leader Election via FHE with Reduced Computation and Communication and Transparent Setup Authors: Haofei Liang (SJTU), Zeyu Liu (Yale), Yunhao Wang (Yale), Xiang Xie (Primus Labs), Yu Yu (SJTU), Fan Zhang (Yale, IC3) ## About Relect is a single secret leader election (SSLE) protocol based on the Ring Learning with Errors (RLWE) assumption. It uses lattice-based FHE building blocks to elect one leader among G parties without revealing the leader's identity. Compared to the prior state-of-the-art Qelect, Relect removes the trusted setup requirement while achieving substantially faster computation and smaller communication. ## Contents - ssle_core/ — main protocol library (keygen, parameters, protocol helpers)- ssle.rs — full distributed election with real TCP networking- ssle_compute_time.rs — FHE compute-time benchmark (G ≤ 128)- ssle_ge_256_compute_time_improve.rs — FHE benchmark (G ≥ 256, 128 RGSWs)- run_bench.sh / run_bench.ps1 — automated benchmark scripts- analyze_bench.sh / analyze_bench.ps1 — result analysis scripts ## Requirements - Rust stable ≥ 1.95.0 + nightly ≥ 1.97.0- Linux (Ubuntu), Windows 11, or macOS- Multi-core CPU recommended for parallel benchmarks ## Quick Start bashrustup install nightly # Correctness test (G = 4 parties)cargo run --release --package ssle_core --example ssle -- -p 4 # FHE benchmark (G = 4)cargo run --release --package ssle_core --example ssle_compute_time -- -p 4 ## Reproducing Paper Results bashbash run_bench.sh -c nightly -s -r 5bash analyze_bench.sh results/<file> --stats See README.md in the archive for full documentation.