mirror of
https://github.com/lux-apps/papers.git
synced 2026-10-30 18:15:21 +00:00
zeekay
abf99d05fc
Restructured from flat .tex/.pdf files into per-paper subdirectories. Added Antje Worring as co-author on all papers. Synced PDFs to pdfs/ for download site.
81 lines
2.5 KiB
TeX
81 lines
2.5 KiB
TeX
\section{Benchmarks}
|
|
|
|
\label{sec:benchmarks}
|
|
|
|
All measurements on AMD EPYC 7763 nodes connected via 1\,Gbps links with
|
|
$\sim$1\,ms round-trip latency between parties. Go~1.22 implementation using
|
|
\texttt{luxfi/threshold} and \texttt{luxfi/mpc} libraries.
|
|
|
|
\subsection{DKG Latency}
|
|
|
|
\begin{table}[ht]
|
|
\centering
|
|
\caption{DKG protocol latency.}
|
|
\label{tab:dkg-benchmarks}
|
|
\begin{tabular}{rrrr}
|
|
\toprule
|
|
\textbf{$(t, n)$} & \textbf{DKG Time} & \textbf{Share Size} & \textbf{Comms} \\
|
|
\midrule
|
|
$(2, 3)$ & 45\,ms & 32\,B & 192\,B \\
|
|
$(3, 5)$ & 112\,ms & 32\,B & 640\,B \\
|
|
$(5, 7)$ & 238\,ms & 32\,B & 1{,}568\,B \\
|
|
$(7, 10)$ & 520\,ms & 32\,B & 3{,}200\,B \\
|
|
$(15, 21)$ & 1.8\,s & 32\,B & 14{,}112\,B \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\end{table}
|
|
|
|
\subsection{Signing Latency}
|
|
|
|
\begin{table}[ht]
|
|
\centering
|
|
\caption{Signing latency by protocol and threshold.}
|
|
\label{tab:signing-benchmarks}
|
|
\begin{tabular}{llrr}
|
|
\toprule
|
|
\textbf{Protocol} & \textbf{$(t, n)$} & \textbf{Signing Latency} & \textbf{Sig Size} \\
|
|
\midrule
|
|
Threshold BLS & $(2, 3)$ & 1.1\,ms & 48\,B \\
|
|
Threshold BLS & $(4, 5)$ & 2.1\,ms & 48\,B \\
|
|
Threshold BLS & $(7, 10)$ & 4.5\,ms & 48\,B \\
|
|
\midrule
|
|
CGGMP21 (ECDSA) & $(2, 3)$ & 2.8\,ms & 65\,B \\
|
|
CGGMP21 (ECDSA) & $(3, 5)$ & 4.2\,ms & 65\,B \\
|
|
CGGMP21 (ECDSA) & $(7, 10)$ & 18.7\,ms & 65\,B \\
|
|
\midrule
|
|
FROST (Ed25519) & $(2, 3)$ & 1.5\,ms & 64\,B \\
|
|
FROST (Ed25519) & $(3, 5)$ & 2.4\,ms & 64\,B \\
|
|
FROST (Ed25519) & $(5, 7)$ & 6.8\,ms & 64\,B \\
|
|
FROST (Ed25519) & $(7, 10)$ & 12.3\,ms & 64\,B \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\end{table}
|
|
|
|
\subsection{Comparison: MPC Signing vs. Single-Key Signing}
|
|
|
|
\begin{table}[ht]
|
|
\centering
|
|
\caption{Overhead of MPC signing relative to single-key signing.}
|
|
\label{tab:overhead}
|
|
\begin{tabular}{lrrr}
|
|
\toprule
|
|
\textbf{Scheme} & \textbf{Single-Key} & \textbf{MPC $(3,5)$} & \textbf{Overhead} \\
|
|
\midrule
|
|
BLS sign & 28\,$\mu$s & 2.1\,ms & $75\times$ \\
|
|
ECDSA sign & 18\,$\mu$s & 4.2\,ms & $233\times$ \\
|
|
EdDSA sign & 22\,$\mu$s & 2.4\,ms & $109\times$ \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\vspace{0.3em}
|
|
\footnotesize{Overhead is dominated by network round-trips, not computation.
|
|
With co-located parties (LAN), overhead drops to $5\text{--}15\times$.}
|
|
\end{table}
|
|
|
|
The absolute latencies (2--19\,ms) are well within the requirements of blockchain
|
|
consensus, which operates on 100\,ms--2\,s block times. MPC signing adds negligible
|
|
latency relative to consensus rounds.
|
|
|
|
%% ========================================================================
|
|
%% 10. RELATED WORK
|
|
%% ========================================================================
|