PQ Crypto Registry

Methodology

How the registry evaluates post-quantum cryptographic algorithms. Each algorithm is assessed across standardization maturity, security properties, risk profile, performance, and implementation readiness.

Scope

The registry tracks post-quantum cryptographic algorithms relevant to practical deployment. Algorithms are included based on standardization progress, adoption potential, or significance to specific ecosystems (e.g. blockchain).

6

Digital signature schemes

2

Key encapsulation mechanisms

Assumption Families

Every algorithm is classified by the mathematical hardness assumption it relies on. This determines the algorithm's fundamental security posture and informs the assumption risk rating.

Lattice
3 algorithms

Security relies on the hardness of problems in high-dimensional lattices, such as finding short vectors (SVP) or distinguishing noisy linear equations (LWE). Offers strong performance and compact keys, with decades of cryptanalytic study.

Hash-based
4 algorithms

Security relies solely on the properties of cryptographic hash functions (collision resistance, preimage resistance). Considered the most conservative assumption family because no structural algebraic assumption is required beyond hash function security.

Code-based
1 algorithm

Security relies on the hardness of decoding random linear error-correcting codes. The McEliece cryptosystem (1978) is the oldest post-quantum proposal in this family, providing strong confidence through decades of study.

Isogeny-based
0 algorithms

Security relies on the hardness of computing isogenies between elliptic curves. A newer family offering compact key sizes but with less mature cryptanalytic history. The SIDH attack in 2022 demonstrated the importance of ongoing analysis.

Standardization Status

The registry records standardization progress from recognized bodies including NIST, IETF, and relevant ecosystem organizations. Each algorithm's status reflects the maturity of its formal specification.

Standard
Published as a final standard by a recognized body (e.g. NIST FIPS, IETF RFC). Suitable for production deployment.
Draft
Standard is in development with a designated reference number. The specification is largely stable but may still undergo minor revisions before finalization.
Candidate
Selected for standardization but the formal specification has not yet been published. Subject to change during the standardization process.
Research
Active academic proposal or early-stage design. Not yet adopted by a standardization body. Included in the registry for tracking and comparison.

Security Properties

Each algorithm is characterized by four security properties that describe its formal guarantees and operational requirements.

Hardness Assumption
The specific mathematical problem that must remain hard for the scheme to be secure (e.g. Module-LWE, QCSD). Distinct from the broader assumption family.
Security Notion
The formal security guarantee the scheme provides. Digital signatures target EUF-CMA (existential unforgeability under chosen message attack). KEMs target IND-CCA2 (indistinguishability under adaptive chosen ciphertext attack).
Deterministic
Whether the core operation (signing or encapsulation) produces the same output given the same inputs. Non-deterministic schemes use internal randomness, which can provide hedging against fault attacks but requires a reliable entropy source.
Statefulness
Whether the signer must maintain state between operations. Stateful schemes (e.g. XMSS) require careful index tracking — reusing state can be catastrophic. Stateless schemes are simpler to deploy but may have larger signatures.

Risk Assessment

Risk assessments use a conservative three-tier scale across three independent dimensions. Definitions are intentionally narrow to avoid ambiguity. Ratings are assigned based on published cryptanalysis, implementation track record, and inherent algorithmic properties.

Assumption Risk

Confidence in the underlying mathematical hardness assumption.

low
Well-studied assumption with decades of scrutiny and no known sub-exponential attacks (classical or quantum). Examples: Module-LWE at standard parameters.
medium
Assumption is broadly believed to hold, but has less cryptanalytic history or relies on structured variants that narrow the safety margin.
high
Assumption is novel, has limited independent analysis, or recent results have weakened confidence in related problems.
Implementation Risk

Difficulty of producing a correct, standards-conformant implementation.

low
Straightforward to implement correctly. Reference code is mature, few edge cases, and misuse-resistant API design.
medium
Requires careful attention to parameter encoding, rejection sampling, or key validation. Subtle bugs have appeared in early implementations.
high
Complex state management, fragile failure modes, or a history of implementation-level vulnerabilities across multiple libraries.
Side-Channel Risk

Susceptibility to timing, power, or electromagnetic side-channel attacks.

low
Core operations are naturally constant-time or trivially hardened. No known exploitable leakage in standard deployment models.
medium
Constant-time implementation is achievable but requires explicit effort (e.g., masked arithmetic, constant-time rejection sampling). Non-hardened code is vulnerable.
high
Inherent algorithmic features (e.g., variable-weight operations, secret-dependent branching) make side-channel resistance difficult even for expert implementers.

Performance Evaluation

Performance is captured at two levels of detail. All algorithms receive qualitative relative ratings. When available, quantitative benchmarks are included with platform and source attribution.

Relative Ratings

Each core operation is rated on a three-point scale relative to other algorithms of the same primitive type.

Fast
Completes in time comparable to or better than classical counterparts for typical parameter sets.
Medium
Noticeably slower than classical counterparts but acceptable for most deployments.
Slow
Significantly slower than classical counterparts. May require architectural consideration for latency-sensitive applications.

Digital signatures are rated on keygen, sign, and verify. KEMs are rated on keygen, encaps, and decaps.

Benchmarks

When available, quantitative benchmarks report wall-clock timing in microseconds for each operation and parameter set. Each benchmark entry includes the hardware platform and source implementation to ensure reproducibility. Benchmarks are not available for all algorithms.

Parameter Sets & Sizes

Each algorithm defines one or more parameter sets at different NIST security levels. The registry records the concrete byte sizes for all cryptographic artifacts.

NIST Security Levels

NIST security level classifications define the minimum computational effort required to break a scheme, benchmarked against symmetric primitives.

Level 1At least as hard to break as AES-128 key recovery
Level 2At least as hard to break as SHA-256 collision
Level 3At least as hard to break as AES-192 key recovery
Level 4At least as hard to break as SHA-384 collision
Level 5At least as hard to break as AES-256 key recovery
Tracked Sizes

Digital Signatures

  • Public key (pk_bytes)
  • Secret key (sk_bytes)
  • Signature (sig_bytes)

KEMs

  • Public key (pk_bytes)
  • Secret key (sk_bytes)
  • Ciphertext (ct_bytes)
  • Shared secret (ss_bytes)

Capabilities

Capabilities describe functional properties beyond core cryptographic operations. These help identify algorithms suited to specific deployment scenarios.

batch-verification
Multiple signatures can be verified simultaneously faster than verifying each individually.
threshold-signatures
The signing key can be split among multiple parties, requiring a threshold of participants to produce a valid signature.
aggregation
Multiple signatures can be combined into a single compact proof of validity.
hedged-signing
Signing incorporates both deterministic and randomized components, providing resilience against both fault injection and poor randomness.
hardware-friendly
Operations map efficiently to constrained hardware (HSMs, smart cards, embedded devices) without requiring large memory or complex arithmetic.
hybrid-mode
Can be deployed alongside a classical algorithm (e.g. ECDSA + ML-DSA) in a hybrid construction for defense-in-depth during migration.
forward-secrecy
Compromise of the current key does not allow decryption of past sessions or forgery of past signatures.
key-agreement
Can be used to establish a shared secret between two parties for symmetric encryption.
snark-aggregation
Signatures are designed for efficient aggregation via STARK/SNARK proofs, enabling scalable on-chain verification.

Implementation Tracking

The registry catalogs known implementations for each algorithm, recording the implementation name, language, source link, and whether the code has undergone a formal security audit.

Audited
The implementation has been reviewed by an independent security auditor. This does not guarantee the absence of vulnerabilities, but indicates a higher level of scrutiny.
Unaudited
No known formal audit. The implementation may still be high quality, but has not been independently verified. Exercise additional caution for production use.

Contributing

The PQ Crypto Registry is open source. Algorithm data, risk assessments, and methodology are maintained in structured YAML files with automated validation. Contributions are welcome via pull request.