mirror of
https://github.com/lux-apps/papers.git
synced 2026-10-30 18:15:21 +00:00
zooqueen
ed0d9cabdf
Scrub residual Ring-LWE mislabels of the Corona/Pulsar threshold
signature family per canonical decision-b (source: ringtail/sign/config.go
M=8,N=7, q=0x1000000004A01).
- Corona = Module-LWE (Ringtail-derived parameter set, A in R_q^{8x7}),
not Ring-LWE / not "Module-LWE at rank 1".
- Pulsar = Module-LWE (ML-DSA-65, FIPS-204 byte-equal).
- Corona+Pulsar differ by parameter regime/codebase, NOT hardness family:
intra-lattice parameter/implementation diversity, NOT family-disjoint.
- True family-disjoint defense = Magnetar (hash, SLH-DSA) / HQC (codes).
Reframed the false Corona(R-LWE)-vs-Pulsar(M-LWE) family-independence
across lp-219/lp-220/lp-222/lp-223 (P3Q kinds), lux-quasar-composition
(profile hardness table + composition corollary/proof), lux-magnetar-pq,
and the superseded triple-proof/quasar drafts (anonymous lattice
threshold lane = Corona). Legitimate FHE/TFHE Ring-LWE, NTT citations,
and neutral "rank-1" lattice facts left intact. PDFs rebuilt.
62 lines
1.7 KiB
TeX
62 lines
1.7 KiB
TeX
\section{Performance}
|
|
|
|
|
|
\subsection{Benchmarks}
|
|
|
|
Measured on Apple M1 Max (10 CPU cores, darwin/arm64, Go~1.26.1).
|
|
GPU numbers are estimated; all others are \texttt{go~test~-bench} output~\cite{lux-benchmarks}.
|
|
|
|
\begin{center}
|
|
\begin{tabular}{lr}
|
|
\toprule
|
|
\textbf{Operation} & \textbf{Time} \\
|
|
\midrule
|
|
BLS sign & 350\,$\mu$s \\
|
|
BLS verify (single) & 820\,$\mu$s \\
|
|
BLS aggregated verify ($N$=100) & 875\,$\mu$s \\
|
|
ML-DSA-65 sign & 495\,$\mu$s \\
|
|
ML-DSA-65 verify (raw) & 181\,$\mu$s \\
|
|
ML-DSA-65 verify (via fx) & 254\,$\mu$s \\
|
|
ML-DSA-65 verify (cached) & 3\,$\mu$s \\
|
|
Groth16 prover (CPU) & ${\sim}400$\,ms \\
|
|
Groth16 prover (GPU, est.) & ${\sim}5$--$15$\,ms \\
|
|
Groth16 verify (3 pairings, CPU) & ${\sim}1$--$3$\,ms \\
|
|
Quasar block processing (BLS+ML-DSA+Corona) & 1.85\,ms \\
|
|
\textbf{QuasarCert total CPU verify} & \textbf{${\sim}2$--$5$\,ms} \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\end{center}
|
|
|
|
\subsection{Storage}
|
|
|
|
At 1ms block times (1,000 blocks/second):
|
|
|
|
\begin{center}
|
|
\begin{tabular}{lrrr}
|
|
\toprule
|
|
\textbf{Component} & \textbf{Size} & \textbf{Frequency} & \textbf{Annual} \\
|
|
\midrule
|
|
Block header & 144 B & 1,000/s & 4.5 TB \\
|
|
Epoch proof & 258 B & 1/s & 8.1 GB \\
|
|
\midrule
|
|
\textbf{Total (without tx data)} & & & \textbf{$\sim$4.5 TB} \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\end{center}
|
|
|
|
With pruning (retain only epoch proofs after 30 days):
|
|
|
|
\begin{center}
|
|
\begin{tabular}{lrrr}
|
|
\toprule
|
|
\textbf{Component} & \textbf{Retained} & \textbf{Annual} \\
|
|
\midrule
|
|
Recent blocks (30 days) & 144 B $\times$ 1000/s $\times$ 30d & 373 GB \\
|
|
Epoch proofs (permanent) & 258 B $\times$ 1/s $\times$ 365d & 8.1 GB \\
|
|
\midrule
|
|
\textbf{Total (pruned)} & & \textbf{$\sim$381 GB} \\
|
|
\bottomrule
|
|
\end{tabular}
|
|
\end{center}
|
|
|