Public API accepts non-canonical expanded ML-KEM decapsulation key representations and exposes inconsistent handling of the resulting state
Los mantenedores suelen responder en 1 día
Nadie ha tomado este issue todavía.
Evaluación
- Dificultad
- 4/5
- Tiempo estimado
- 3-5 días
- Aptitud para principiantes
- 55/100
- Tipo de issue
- Error
- Claridad
- Bastante claro
- Estado de actividad
- Activo
- Stack tecnológico
- rust
- Área
- api, cryptography, security
Línea de trabajo
Start by reading decapsulation_key.rs at from_expanded(), PartialEq, and KeyExport::to_bytes(), then compare the validation path in pke.rs with the error handling in pkcs8.rs. Reproduce the non-canonical expanded-key and export cases, add regression coverage, and make the chosen validation and failure behavior consistent with the project’s intended API contract.
Escrito por el modelo de indexación a partir del texto del issue.
Descripción
Public API accepts non-canonical expanded ML-KEM decapsulation key representations and exposes inconsistent handling of the resulting state
Summary
DecapsulationKey::from_expanded() (public, #[deprecated]) and its trait
wrapper ExpandedKeyEncoding::from_expanded_bytes() (line 272, delegates at
line 273) accept non-canonical 12-bit coefficient encodings in the decryption
key region without validation. Both are public APIs — from_expanded_bytes()
delegates directly to from_expanded(), which in turn calls the crate-internal
DecryptionKey::from_bytes() (infallible, no modulus check).
The accepted representation produces an internal state (d = None, line 77)
that is observably different from seed-constructed keys in one specific
capability (to_seed), despite PartialEq reporting equality (comment at
line 131, implementation at line 137 — intentionally omits d). This state
divergence propagates to KeyExport::to_bytes() as a panic (line 232), while
the same condition is handled gracefully as an error in the PKCS#8 path
(line 162).
No cryptographic divergence was observed: 546 non-canonical representations
across ML-KEM-512, ML-KEM-768, and ML-KEM-1024 all produced identical shared
secrets compared to their canonical equivalents.
All line references are against commit 1d535a1178476a55b308efb510f86bdf7fd5330b.
Tested revision
Commit: 1d535a1178476a55b308efb510f86bdf7fd5330b
Environment
- rustc 1.96.0 (ac68faa20 2026-05-25)
- cargo 1.96.0 (30a34c682 2026-05-25)
- Linux 6.17.0-40-generic x86_64
Observed properties
1. Non-canonical dk acceptance (representation)
The public entry points are DecapsulationKey::from_expanded() (line 64,
pub, #[deprecated]) and ExpandedKeyEncoding::from_expanded_bytes()
(line 272, pub trait method, delegates to from_expanded() at line 273).
Both are compilable — the deprecation notice is the only signal to consumers.
from_expanded() calls the crate-internal DecryptionKey::from_bytes() in
pke.rs (line 111). This function is infallible — it calls decode_u12,
which applies byte_decode with a 12-bit mask (& 0x0FFF) followed by
small_reduce (conditional subtraction of Q=3329). Coefficients in
[3329, 4095] are silently reduced to [0, 766].
This contrasts with EncryptionKey::from_bytes() (line 168), which performs
the FIPS 203 §7.2 modulus check via encode-decode round-trip and returns
InvalidKey on mismatch.
Note: from_expanded() does validate the encryption key portion of the
expanded representation (via EncryptionKey::from_bytes()?) and checks the
hash (ek.h() != *h). The decryption key portion is the one accepted without
validation.
Reproduction: Modify any dk_pke coefficient c < 767 to c + 3329 in an
expanded key. from_expanded_bytes() accepts it. Re-export via
to_expanded_bytes() produces the original canonical form.
Falsification: 195/195 non-canonical coefficients accepted (ML-KEM-768,
single key). 546/546 non-canonical representations across all three parameter
sets produced identical shared secrets.
2. Panic in KeyExport::to_bytes() (availability)
from_expanded() sets d = None (line 77 of decapsulation_key.rs).
KeyExport::to_bytes() calls
self.to_seed().expect("should be initialized from a seed") (line 232),
which panics.
The same condition is handled in pkcs8.rs (line 162) via
self.to_seed().ok_or(pkcs8::KeyError::Invalid)?.
Reproduction:
#[allow(deprecated)]
let dk = DecapsulationKey::from_expanded_bytes(&expanded).unwrap();
let _seed: Seed = dk.to_bytes(); // panics at decapsulation_key.rs:232:24
3. PartialEq hides capability divergence (API contract)
The PartialEq implementation intentionally omits d from comparison
(comment at line 131, implementation at line 137 of decapsulation_key.rs).
Two keys that compare equal may have different capabilities:
| Operation | d = Some (from_seed) |
d = None (from_expanded) |
|---|---|---|
decapsulate |
works | works (identical result) |
encapsulation_key |
works | works (identical result) |
to_expanded_bytes |
works | works (identical result) |
to_seed() |
Some(seed) |
None |
KeyExport::to_bytes() |
returns seed | panics |
PKCS#8 to_pkcs8_der() |
works | returns KeyError::Invalid |
Security characterization
No cryptographic divergence was observed. The demonstrated impact is:
- A public-API panic (denial of service) reachable through the deprecated but
compilablefrom_expandedpath followed byKeyExport::to_bytes(). - An inconsistent error/panic contract for the same internal state across
KeyExportvs PKCS#8. - An equality contract that does not reflect functional equivalence.
Normative note
FIPS 203 §7.2 specifies a modulus check for encapsulation keys (implemented).
The corresponding validation requirements for expanded decapsulation key
representations under §7.3 remain under review by the reporter. This issue
does not assert a FIPS 203 conformance gap — it reports the observed behavior
and its consequences.
Suggested remediation
- Replace
.expect()inKeyExport::to_bytes()(line 232) with error
handling consistent with the PKCS#8 path (e.g., returnOption<Seed>
or propagate an error). - Consider adding a modulus check to
DecryptionKey::from_bytes()analogous
to theEncryptionKeyround-trip validation, or document the asymmetry
as an intentional design decision. - Consider documenting the
PartialEqsemantics regardingdor providing
anfn is_seed_exportable(&self) -> boolmethod.
- Lenguaje dominante
- Rust
- Estrellas
- 118
- Forks
- 58
- Merge medio
- 29 min
- PR fusionados (30 d)
- 2
Preparar el entorno
Este proyecto no incluye contenedor de desarrollo, Dockerfile ni guía de contribución, así que la configuración corre por tu cuenta: empieza por su README y consulta nuestra guía para la primera contribución para los pasos generales.
Primeros pasos
- Lee el issue completo y luego la guía de contribución del proyecto.
- Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
- Haz un fork del repositorio y trabaja en una rama.
- Abre un pull request que haga referencia al número del issue.
Más de RustCrypto/KEMs
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 35/100
RustCrypto/KEMs#395 ·
Los mantenedores suelen responder en 1 día
-
frodo-kem: `Algorithm::encryption_key_from_decryption_key` skips both of the validations its siblings performPosiblemente ocupada @cipherprofessor la tomó hace 25 días. Abierto
Dificultad 4/5 3-5 días Aptitud para principiantes 58/100
RustCrypto/KEMs#381 ·
Los mantenedores suelen responder en 1 día
-
Dificultad 5/5 Más de una semana Aptitud para principiantes 20/100
RustCrypto/KEMs#346 · 2 comentarios ·
Los mantenedores suelen responder en 1 día
-
Dificultad 5/5 Más de una semana Aptitud para principiantes 20/100
RustCrypto/KEMs#328 ·
Los mantenedores suelen responder en 1 día
-
Dificultad 5/5 Más de una semana Aptitud para principiantes 30/100
RustCrypto/KEMs#325 · 1 comentario ·
Los mantenedores suelen responder en 1 día
Todos los issues de RustCrypto/KEMs
Issues similares
-
bug
Dificultad 2/5 1-3 horas Aptitud para principiantes 72/100
Los mantenedores suelen responder en 1 día
-
needs: triage type: bug
Dificultad 2/5 1-3 horas Aptitud para principiantes 70/100
Los mantenedores suelen responder en 1 día
-
area:docs documentation P3
Dificultad 1/5 Menos de una hora Aptitud para principiantes 88/100
Raaif-Yousuf/Wingman#837 ·
Los mantenedores suelen responder en 1 día
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 80/100
Los mantenedores suelen responder en 1 día
-
priority: medium
Dificultad 2/5 1-3 horas Aptitud para principiantes 68/100
storytold/photocraft#6759 ·
Los mantenedores suelen responder en 1 día