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.
78 lines
2.5 KiB
TeX
78 lines
2.5 KiB
TeX
\section{Performance Evaluation}
|
|
|
|
|
|
\subsection{Experimental Setup}
|
|
|
|
\paragraph{Hardware}
|
|
\begin{itemize}
|
|
\item CPU: Apple M1 Pro / Intel i7-12700K
|
|
\item RAM: 32GB DDR4
|
|
\item Network: Localhost (0.1ms latency) / LAN (5ms latency)
|
|
\end{itemize}
|
|
|
|
\paragraph{Implementation} Go 1.24.5, constant-time arithmetic via saferith library, parallel processing for heavy computations.
|
|
|
|
\subsection{Benchmark Results}
|
|
|
|
\begin{table*}[t]
|
|
\centering
|
|
\begin{tabular}{@{}lccccc@{}}
|
|
\toprule
|
|
\textbf{Operation} & \textbf{3-of-5} & \textbf{5-of-9} & \textbf{7-of-11} & \textbf{10-of-15} & \textbf{Complexity} \\
|
|
\midrule
|
|
\textbf{CMP Keygen} & 12ms & 28ms & 45ms & 82ms & $O(n^2)$ \\
|
|
\textbf{CMP Presign} & 45ms & 98ms & 152ms & 230ms & $O(n^2)$ \\
|
|
\textbf{CMP Sign} & 15ms & 32ms & 48ms & 75ms & $O(n)$ \\
|
|
\midrule
|
|
\textbf{FROST Keygen} & 10ms & 25ms & 42ms & 78ms & $O(n^2)$ \\
|
|
\textbf{FROST Sign} & 8ms & 18ms & 28ms & 45ms & $O(n)$ \\
|
|
\midrule
|
|
\textbf{LSS Keygen} & 12ms & 28ms & 45ms & 82ms & $O(n^2)$ \\
|
|
\textbf{LSS Reshare} & 20ms & 35ms & 52ms & 75ms & $O(n \cdot n')$ \\
|
|
\textbf{LSS Sign} & 18ms & 35ms & 55ms & 88ms & $O(n)$ \\
|
|
\midrule
|
|
\textbf{Doerner Keygen} & 8ms & -- & -- & -- & $O(1)$ (2-party) \\
|
|
\textbf{Doerner Sign} & 5ms & -- & -- & -- & $O(1)$ (2-party) \\
|
|
\midrule
|
|
\textbf{Verification} & 2ms & 2ms & 2ms & 2ms & $O(1)$ \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
|
|
\label{tab:benchmarks}
|
|
\end{table*}
|
|
|
|
\subsection{Scalability Analysis}
|
|
|
|
\paragraph{Key Generation} Complexity $O(n^2)$ due to pairwise share distribution. Optimizations:
|
|
\begin{itemize}
|
|
\item Parallel proof generation (4x speedup on 4 cores)
|
|
\item Batched commitment verification
|
|
\item Efficient broadcast using Merkle trees
|
|
\end{itemize}
|
|
|
|
\paragraph{Signing} Complexity $O(n)$ for online phase. CMP presigning amortizes costs across multiple signatures.
|
|
|
|
\paragraph{Network Latency Impact}
|
|
\begin{itemize}
|
|
\item Localhost (0.1ms): Minimal impact, CPU-bound
|
|
\item LAN (5ms): 15-25\% overhead for multi-round protocols
|
|
\item WAN (50ms): 2-3x slowdown, dominated by network rounds
|
|
\end{itemize}
|
|
|
|
\subsection{Throughput}
|
|
|
|
\paragraph{Concurrent Signing} Independent signatures can be parallelized:
|
|
\begin{itemize}
|
|
\item CMP: 200 signatures/sec (with presignature pool)
|
|
\item FROST: 350 signatures/sec
|
|
\item Doerner: 800 signatures/sec (2-party)
|
|
\end{itemize}
|
|
|
|
\paragraph{Presignature Generation} CMP presignatures can be precomputed:
|
|
\begin{itemize}
|
|
\item 3-of-5: 22 presignatures/sec
|
|
\item 5-of-9: 10 presignatures/sec
|
|
\item 10-of-15: 4 presignatures/sec
|
|
\end{itemize}
|
|
|