GCVE Workshop - 22 September 2026 (14:00-18:00), Luxembourg Before The Vulnopticon Conference - Registration

GHSA-G2P6-HH5V-7HFM

Vulnerability from github – Published: 2026-03-13 15:40 – Updated: 2026-03-13 15:40
VLAI
Summary
Poseidon V1 variable-length input collision via implicit zero-padding
Details

Impact

Poseidon V1 (PoseidonSponge) accepts variable-length inputs without injective padding. When a caller provides fewer inputs than the sponge rate (inputs.len() < T - 1), unused rate positions are implicitly zero-filled. This allows trivial hash collisions: for any input vector [m1, ..., mk] hashed with a sponge of rate > k, hash([m1, ..., mk]) equals hash([m1, ..., mk, 0]) because both produce identical pre-permutation states.

This affects any use of PoseidonSponge or poseidon_hash where the number of inputs is less than T - 1 (e.g., hashing 1 input with T=3).

Poseidon2 (Poseidon2Sponge) is not affected — it encodes the input length in the capacity element (IV = input_len << 64), making different-length inputs produce distinct states.

Patches

Fixed by enforcing inputs.len() == RATE in PoseidonSponge::compute_hash, matching circom's invariant that nInputs always equals T - 1. Users should upgrade to the next release containing this fix.

Workarounds

If upgrading is not immediately possible:

  • Ensure callers always use T = inputs.len() + 1 (full-rate), which is how circom uses Poseidon. For example, to hash 2 inputs, use T=3; to hash 1 input, use T=2. Never use a sponge with more rate capacity than the number of inputs.
  • Alternatively, migrate to Poseidon2Sponge, which is safe for variable-length inputs due to its length-encoding IV.

References

Severity
8.7 (High)
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:H/VA:N/SC:N/SI:N/SA:N
Show details on source website
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "crates.io",
        "name": "soroban-poseidon"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "25.0.1"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-32129"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-328"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-03-13T15:40:31Z",
    "nvd_published_at": "2026-03-12T18:16:25Z",
    "severity": "HIGH"
  },
  "details": "## Impact\n\nPoseidon V1 (`PoseidonSponge`) accepts variable-length inputs without injective padding. When a caller provides fewer inputs than the sponge rate (`inputs.len() \u003c T - 1`), unused rate positions are implicitly zero-filled. This allows trivial hash collisions: for any input vector `[m1, ..., mk]` hashed with a sponge of rate \u003e k, `hash([m1, ..., mk])` equals `hash([m1, ..., mk, 0])` because both produce identical pre-permutation states.\n\nThis affects any use of `PoseidonSponge` or `poseidon_hash` where the number of inputs is less than `T - 1` (e.g., hashing 1 input with `T=3`).\n\nPoseidon2 (`Poseidon2Sponge`) is **not affected** \u2014 it encodes the input length in the capacity element (`IV = input_len \u003c\u003c 64`), making different-length inputs produce distinct states.\n\n## Patches\n\nFixed by enforcing `inputs.len() == RATE` in `PoseidonSponge::compute_hash`, matching circom\u0027s invariant that `nInputs` always equals `T - 1`. Users should upgrade to the next release containing this fix.\n\n## Workarounds\n\nIf upgrading is not immediately possible:\n\n- Ensure callers **always** use `T = inputs.len() + 1` (full-rate), which is how circom uses Poseidon. For example, to hash 2 inputs, use `T=3`; to hash 1 input, use `T=2`. Never use a sponge with more rate capacity than the number of inputs.\n- Alternatively, migrate to `Poseidon2Sponge`, which is safe for variable-length inputs due to its length-encoding IV.\n\n## References\n- [circom Poseidon implementation](https://github.com/iden3/circomlib/blob/master/circuits/poseidon.circom) \u2014 reference implementation where `nInputs` determines `T`\n- [Poseidon paper](https://eprint.iacr.org/2019/458) \u2014 Section 4 discusses sponge construction and padding requirements",
  "id": "GHSA-g2p6-hh5v-7hfm",
  "modified": "2026-03-13T15:40:31Z",
  "published": "2026-03-13T15:40:31Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/stellar/rs-soroban-poseidon/security/advisories/GHSA-g2p6-hh5v-7hfm"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-32129"
    },
    {
      "type": "WEB",
      "url": "https://github.com/stellar/rs-soroban-poseidon/pull/10"
    },
    {
      "type": "WEB",
      "url": "https://github.com/stellar/rs-soroban-poseidon/commit/ceb20d3593fc4a8a951a7e99d8fa2344f8250a8c"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/stellar/rs-soroban-poseidon"
    },
    {
      "type": "WEB",
      "url": "https://github.com/stellar/rs-soroban-poseidon/releases/tag/v25.0.1"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:H/VA:N/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "Poseidon V1 variable-length input collision via implicit zero-padding"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.





Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.


Detection rules are retrieved from Rulezet.


Related by attack behaviour

Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.