NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
pub.dev
pub.dev, #1963 most downloaded on pub.dev
Pure Dart post-quantum cryptography. Zero dependencies. ML-KEM, ML-DSA, and SLH-DSA with checked-in NIST vector evidence.
Latest stable
0.4.4
Last release 24 days ago
29 Oct 2026
Release timing varies
gaps range from 2 weeks to 6 months
Nearly every release is documented
notes for 10 of 10 stable releases
Nothing withdrawn
no release was ever pulled
11 months old
10 releases · first in 2025
Drag across the ruler to compare, or tap a release to open it.
Verified by the release workflow:
pqcrypto v0.4.4
Verified by the release workflow:
dart formatdart analyzetool/check_publish_surface.dartdart doctest/kat_evaluator_test.dart)test/mldsa_kat_test.dart)dart pub publish --dry-runpubspec.yaml versionSLH-DSA ACVP (all 12 sets) and dart2js/dart2wasm stay on CI.
pub.dev publishing is automatic via
.github/workflows/publish.yml
(GitHub Actions OIDC, environment pub.dev) on vX.Y.Z tags.
This is not a FIPS 140 module. See doc/FIPS_140_BOUNDARY.md.
Full Changelog: https://github.com/turkananation/pqcrypto/compare/v0.4.3...v0.4.4
SlhDsa.generateKeyPairSeeded (FIPS 205 Algorithm 18,
slh_keygen_internal): derives an SLH-DSA key pair from caller-supplied
SK.seed, SK.prf and PK.seed, mirroring MlDsa.generateKeyPairSeeded.
It lets callers validate an imported SLH-DSA secret key: PK.root is fully
determined by SK.seed and PK.seed, so recomputing it detects a corrupted
or inconsistent key that would otherwise import silently and produce only
signatures that fail verification. It also allows deterministic key
generation from the caller's own CSPRNG or key derivation.
The algorithm was already implemented and ACVP-validated as the
source-only SlhDsaInternal.generateKeyPairSeeded, which now delegates to
the public method. No behaviour change.
Your coding agent can read these notes before it upgrades. Set up the MCP server →
Verified by the release workflow:
pqcrypto v0.4.3
Verified by the release workflow:
dart formatdart analyzetool/check_publish_surface.dartdart doctest/kat_evaluator_test.dart)test/mldsa_kat_test.dart)dart pub publish --dry-runpubspec.yaml versionSLH-DSA ACVP (all 12 sets) and dart2js/dart2wasm stay on CI.
pub.dev publishing is automatic via
.github/workflows/publish.yml
(GitHub Actions OIDC, environment pub.dev) on vX.Y.Z tags.
This is not a FIPS 140 module. See doc/FIPS_140_BOUNDARY.md.
Full Changelog: https://github.com/turkananation/pqcrypto/compare/v0.4.0...v0.4.3
KyberParams and KyberLevel are exported from package:pqcrypto/pqcrypto.dart.
ML-KEM was the only family whose derived sizes were unreachable. DilithiumParams,
DilithiumParameter, SlhDsaParams and SlhDsaParameter were all exported;
the ML-KEM equivalents were not, because the barrel clause on kyber/kem.dart
was show KyberKem, PqcKem.
That forced any caller sizing an ML-KEM buffer to hardcode literals, which is exactly the drift the exported tables exist to prevent — and it is why a downstream package carried a hand-copied key-length table.
No behaviour change. KyberParams has existed and been used internally; this
only makes it reachable.
A packaging and publishing release. The cryptography did not change. ML-KEM, ML-DSA, and all 12 SLH-DSA sets are the same byte-exact implementations a
A packaging and publishing release. The cryptography did not change. ML-KEM, ML-DSA, and all 12 SLH-DSA sets are the same byte-exact implementations as 0.4.1.
pub.dev scored 0.4.1 at 150/160. The missing ten points were documentation: dart doc crashed while initializing categories.
dartdoc_options.yaml declared a Cookbook category pointing at doc/cookbook/README.md. The 0.4.1 tarball shipped the yaml and omitted the markdown. pana then failed with:
In categories definition for Cookbook, "markdown" resolves to the missing file
.../doc/cookbook/README.md
Dependents that import package:pqcrypto (including pqforge) inherited the same crash. That is #60.
doc/ tree is in the pub.dev archive, including doc/cookbook/README.md.dart pub publish --dry-run and that dart doc succeeds (~46% of the public API is documented, above the 20% bar).vX.Y.Z tag runs GitHub Actions OIDC to pub.dev (publish.yml, environment pub.dev) and opens this GitHub Release (release.yml). Do not also dart pub publish by hand.dependencies:
pqcrypto: ^0.4.2No API or wire-format changes. pub upgrade is enough.
This is KAT / ACVP / interop evidence. It is not a CMVP / FIPS 140 module. See doc/FIPS_140_BOUNDARY.md.
Full changelog: v0.4.1...v0.4.2
doc/ in .pubignore so the Cookbook markdown
(doc/cookbook/README.md) and the rest of the docs tree ship in the
pub.dev archive. dartdoc_options.yaml already declared the Cookbook
category; 0.4.1 omitted the markdown file, so dart doc crashed while
initializing categories (and dependents such as pqforge inherited the
crash). CI now asserts the file is in dart pub publish --dry-run and
that dart doc succeeds. No cryptographic or API changes.vX.Y.Z runs
.github/workflows/publish.yml (OIDC, environment pub.dev) and
.github/workflows/release.yml (verify + GitHub Release). Same layout as
pqforge and pqtransport. This is a library — no CLI binary matrix.docs: define release branch flow by @turkananation in https://github.com/turkananation/pqcrypto/pull/35
Full Changelog: https://github.com/turkananation/pqcrypto/compare/v0.3.1...v0.4.1
Full Changelog: v0.3.1...v0.4.1
example/main.dart into a full-family walkthrough for ML-KEM,
ML-DSA, SLH-DSA, and a signed ML-KEM + ML-DSA handshake composition.Hmac, Keccak, KeccakParameters, Mgf1, Sha2, Shake, Zeroize) at
the package root for cookbook and framework use.Added FIPS 205 SLH-DSA for all 12 parameter sets — both hash families (SHAKE and SHA-2) across 128/192/256 and the small/fast s / f variants — exporte
s/f variants —SlhDsa, SlhDsaParams, SlhDsaParameter,SlhDsaPreHash. Internal Algorithms 18-20 stay source-only for ACVPADRS^c) and the security-category 1 vs 3/5BigInt tree indices, hedged-by-default signing, explicitallowSlowSigning) paths, context binding0.4.0, and the generated website,Updated pubspec.yaml description to accurately reflect full FIPS 204 ML-DSA support, removing the "experimental" label, and highlighting the zero-depe
Updated pubspec.yaml description to accurately reflect full FIPS 204 ML-DSA support, removing the "experimental" label, and highlighting the zero-dependency architecture.
No code changes.
Full FIPS 204-aligned ML-DSA digital signature support:
Full FIPS 204-aligned ML-DSA digital signature support:
sign/verify with context strings, hedged-by-default signinghashSign/hashVerify with FIPS 204 §5.4 SHA-2 pre-hashfinally blocksexample/main.dart now demonstrates ML-KEM + ML-DSA handshakedoc/SERVERPOD_FLUTTER_GUIDE.md for Serverpod/Flutter integrationRejBoundedPoly (ExpandS) for η=2: correct half-byte acceptance per FIPS 204, fixing key generation divergencedart2jsThis release makes no CMVP/FIPS 140 module validation claim. Conformance evidence is the checked-in KAT corpus and regression suite. See doc/FIPS_140_BOUNDARY.md for details.
Install: dart pub add pqcrypto or add pqcrypto: ^0.3.0 to your pubspec.yaml
pub.dev: https://pub.dev/packages/pqcrypto
doc/UNIVERSAL_MULTI_AGENT_PQC_FRAMEWORK.md;tool/agent_framework/pqc_framework.yaml;example/main.dart now demonstrates ML-KEM shared-secret agreement,
ML-DSA signing/verification, and an ML-DSA-signed ML-KEM-768 handshake
transcript.doc/SERVERPOD_FLUTTER_GUIDE.md now covers ML-KEM + ML-DSA Serverpod/Flutter
integration, strict byte contracts, generated model sketches, Flutter isolate
guidance, and framework-driven implementation prompts.deterministic, hedged) × implementation flavour (raw/internal,
pure/external-with-context, hashed/HashML-DSA): 300/300 key generations
and 1800/1800 signatures reproduced byte-for-byte, all verifying.MlDsa: generateKeyPair(params) (fresh
randomness), sign/verify with a context string (≤ 255 bytes) and
hedged-by-default signing, signDeterministic, and the internal/CAVP
helpers generateKeyPairSeeded, signInternal, verifyInternal.hashSign/hashVerify) with the FIPS 204 §5.4 pre-hash:
SHA-256 (ML-DSA-44), SHA-384 (ML-DSA-65), SHA-512 (ML-DSA-87), with DER OID
domain separation.lib/src/common/sha2.dart): SHA-256/384/512,
web-safe 64-bit (hi/lo pair) arithmetic, pinned by direct NIST vectors.test/data/MLDSA (18 .rsp files) with a
discovered, deterministic runner test/mldsa_kat_test.dart; ML-KEM corpus
moved to test/data/MLKEM.dart2js, and dart2wasm.SECURITY.md (vulnerability reporting / coordinated disclosure policy) and
doc/FIPS_140_BOUNDARY.md (why algorithm conformance is not CMVP/FIPS 140
module validation), linked from every place the claim boundary is raised.ML-KEM decapsulation hardening: the FIPS 203 output is now selected with a
constant-time branchless mask — both K' and J(z || c) are always computed,
so success vs. implicit-rejection no longer differs in control flow. Secret
intermediates (m', K' || r', J(z || c), c', z) are zeroized in a
finally block, and the KEM now reuses a cached Random.secure(). The
3000-vector ML-KEM KAT corpus (including invalid-ciphertext vectors) remains
byte-exact.
ML-DSA rejection samplers (RejNTTPoly, RejBoundedPoly, SampleInBall) now
use an incremental SHAKE XOF (KeccakXof) instead of fixed buffers, so
they cannot exhaust output during normal operation.
ML-DSA packing preserves signed coefficient domains for s1/s2/t0/z
instead of folding negatives into [0, q-1].
_normExceeds (the ML-DSA norm gate) evaluates all 256 coefficients with no
secret-dependent early exit, and uses only VM/web-portable arithmetic.
Verification is total: MlDsa.verify/verifyInternal return false
(never throw) for wrong pk/sig lengths, malformed hints, or over-long context.
DilithiumParams exposes computed FIPS 204 Table 2 sizes
(publicKeyBytes, secretKeyBytes, signatureBytes, plus per-poly sizes,
lambda, securityCategory) as the single source of truth; the unused
crhBytes constant was removed.
RejBoundedPoly (ExpandS) for η=2: now accepts half-bytes < 15 mapping
to 2 − (b mod 5) per FIPS 204, fixing the stream-consumption rate that made
key generation diverge from the standard. This was the sole core defect.<< d) that overflowed
on dart2js; it multiplies by 2^d so web results match the VM.lib/src/common/zeroize.dart) applied in
finally blocks around key generation and signing intermediates. Dart cannot
guarantee hard memory erasure; see doc/SECURITY_AUDIT.md for the boundary.Expanded the OpenSSL interoperability harness from ML-KEM-768 only to ML-KEM-512, ML-KEM-768, and ML-KEM-1024.
J(z || c) agreement.dart2js, and dart2wasm.pointycastle runtime dependency by vendoring the FIPS 202
SHA3-256, SHA3-512, SHAKE128, and SHAKE256 implementation in-tree.Input validation for encapsulate() and decapsulate() per FIPS 203 §7.2/§7.3:
encapsulate() and decapsulate() per FIPS 203 §7.2/§7.3:
ByteEncode₁₂ ∘ ByteDecode₁₂ round-trip).H(pk) integrity check.Pack.decodeSecretKey length guard.tool/openssl_interop/): dart:ffi-based harness proving wire-level ML-KEM-768 compatibility with OpenSSL ≥ 3.5. Four-way test matrix (A/B/C/D) validates byte-identical shared secrets across implementations.ci.yml: format check, static analysis, and full test suite (unit + 3000-vector KAT corpus) on every push/PR.interop.yml: builds OpenSSL 4.0.0 from source (cached), runs the four interop tests on every push/PR.kem_validation_test.dart: exercises all input validation paths across ML-KEM-512/768/1024.keygen_derivation_test.dart: isolates FIPS 203 domain separation (G(d || k)) and matrix expansion ordering.poly_test.dart: verifies barrettReduce returns canonical residues in [0, q).doc/MLKEM_TESTING.md: KAT file hashes, coverage boundaries, release-gate commands, and scoped claim boundary.doc/OPENSSL_INTEROP.md: full interop guide with FFI bindings, versions, results, and use cases.genMatrixEntryForTest / sampleNttForTest on Indcpa (internal, not exported)..pubignore to exclude dev-only files from the published package.lowerCamelCase (_H/_G/_J → _h/_g/_j; A_hat/t_hat/r_hat → aHat/tHat/rHat; etc.). No behavioral change.barrettReduce(): use const for compile-time constants and add a fallback res %= q guard for edge-case residues.test/kat_evaluator.dart → test/kat_evaluator_test.dart so dart test discovers it automatically.pubspec.yaml description: fixed typo ("Startss" → "Starts"), updated wording to "FIPS 203-aligned".Poly.montgomeryReduce() (the implementation uses Barrett reduction exclusively).Implements ML-KEM (Kyber) FIPS 203 standard.
pqcrypto.One column per month.