ML-DSA·Digital signature·NIST level 3

ML-DSA-65

ML-DSA (FIPS 204) is a lattice-based digital signature scheme, standardized by NIST in August 2024. Built on Module Learning With Errors, it offers fast verification and moderate signature sizes.

Latest run2026-10-28 11:58 UTC
commitfb7d298
CPU
Xeon Platinum 8488C
Measured on
c7i.large
OS / Kernel
24.04.4 LTS
liboqs
0.15.0
Wall clock
32m 35s
CPU steal
<0.01%

Parameters

Family
ML-DSA
Parameter set
65
Public key
1.91 KB
Private key
3.94 KB
Signature
3.23 KB
Iterations / op
1,000
liboqs key
ML-DSA-65

Regulatory approval status

Whether this algorithm family is approved/recommended for use under three regimes. Every badge links to its source — click through before citing this elsewhere.

keygenGenerate a fresh key pair.
168 of 168 runs
Mean
29.0 µs
95% CI on mean
±0.80%
Median
28.1 µs
p95
35.2 µs
p99
39.6 µs
Stdev
3.7 µs
Min
27.5 µs
Ops/sec
34,483

Mean is known to 28.77–29.23 µs across 1,000 iterations. A 95% confidence interval on the mean, not a prediction interval for a single operation. It says how precisely the average is known from this run's samples; the standard deviation says how much the individual samples scattered.

Mean over time

Click any point to view the GitHub Actions run that produced it.

signProduce a signature over a message using the private key.
168 of 168 runs
Mean
89.3 µs
95% CI on mean
±3.77%
Median
73.1 µs
p95
193.3 µs
p99
282.1 µs
Stdev
54.3 µs
Min
38.8 µs
Ops/sec
11,198

Mean is known to 85.94–92.67 µs across 1,000 iterations. A 95% confidence interval on the mean, not a prediction interval for a single operation. It says how precisely the average is known from this run's samples; the standard deviation says how much the individual samples scattered.

Mean over time

Click any point to view the GitHub Actions run that produced it.

verifyVerify a signature against a message and public key.
168 of 168 runs
Mean
35.1 µs
95% CI on mean
±1.60%
Median
29.6 µs
p95
50.9 µs
p99
64.3 µs
Stdev
9.0 µs
Min
27.9 µs
Ops/sec
28,495

Mean is known to 34.53–35.65 µs across 1,000 iterations. A 95% confidence interval on the mean, not a prediction interval for a single operation. It says how precisely the average is known from this run's samples; the standard deviation says how much the individual samples scattered.

Mean over time

Click any point to view the GitHub Actions run that produced it.

Know the cost of post-quantum migration.

Coldproof