# GHSA-3g92-f9ch-qjcm: Vulnerability Currently, when hashing, if the number of elements to hash is not a multiple of the rate, `hash_iter` pads by elements of the current state. > An attacker can create hash collisions for inputs of different lengths in Plonky3, potentially bypassing integrity checks in cryptographic proofs. - **Severity:** Not scored - **Exploited in the wild:** Not known - **Login needed:** Not needed - **Attackable:** Over the internet - **Victim action needed:** Not needed - **Fix available:** No - **Published:** 16 Apr 2026 ## What is GHSA-3g92-f9ch-qjcm? GHSA-3g92-f9ch-qjcm is a vulnerability in p3-symmetric, published on 16 Apr 2026. An attacker can exploit it over the internet without logging in, with no action from the victim. ### Official description ### Vulnerability Currently, when hashing, if the number of elements to hash is not a multiple of the rate, `hash_iter` pads by elements of the current state. This means that it is possible to create iterators of different lengths which lead to an identical hashed state. Given a simple example using a `PaddingFreeSponge` with width 8 and rate 4. Start with the zero state: [0, 0, 0, 0, 0, 0, 0, 0] Take the first 4 elements to hash and insert into the first 4 elements of the state: [h0, h1, h2, h3, 0, 0, 0, 0] Run the cryptographic permutation on the state: [p00, p10, p20, p30, p40, p50, p60, p70] Take the next 4 elements to hash and insert into the first 4 elements of the state: [h4, h5, h6, h7, p40, p50, p60, p70] Run the cryptographic permutation: [p01, p11, p21, p31, p41, p51, p61, p71] Repeat the above two steps until all elements of the iterator have been consumed. If the number of elements in the iterator is not a multiple of 4 (say there are 10 elements) then, in the final round, the first 2 elements are overwritten and so our final hash would be of: [h8, h9, p21, p31, p41, p51, p61, p71] This means that the iterators over the elements [h0, h1, h2, h3, h4, h5, h6, h7, h8, h9] and [h0, h1, h2, h3, h4, h5, h6, h7, h8, h9, p21] would lead to the same final state of the hasher. ### Impact The impact of this vulnerability is a little difficult to estimate. It is important to note that, in circumstances where the number of elements to be hashed is known and fixed in advance, (as is the case for most STARKS), the method is collision resistant. This vulnerability only applies if a malicious user is able to manipulate the number of elements to be hashed. That being said, there are theoretically situations where this could allow for an amortising of grinding costs (if a prover can manipulate things to get the same hasher state across multiple proofs). ### Patches The fix comes in two parts. The documentation on the current struct `PaddingFreeSponge` has been improved to clarify its intended use case and highlight that it is not collision resistant if an attacker can modify the number of elements being hashed. In addition we add a new struct `Pad10Sponge` which is slightly less efficient but safe in all cases. The padding strategy of the new struct is as follows: If the number of elements in the iterator is not a multiple of the rate, use a 10 padding scheme. If it is a multiple of the rate add 1 to the first secret state element. In the above example, for hashes of length 9, 10, 11, 12, the final state to be permuted would be [h8, 1, 0, 0, p41, p51, p61, p71] [h8, h9, 1, 0, p41, p51, p61, p71] [h8, h9, h10, 1, p41, p51, p61, p71] [h8, h9, h10, h11, p41 + 1, p51, p61, p71] As can be seen, it is now impossible for iterators of different lengths to produce the same "final state" to be hashed which restores collision resistance. (See the following for more details [padding-in-sponge.pdf](https://github.com/user-attachments/files/24465342/padding-in-sponge.pdf)) ### Thanks Many thanks to Benedikt Wagner, Dmitry Khovratovich and Bart Mennink for reporting this issue. ## Which versions are affected by GHSA-3g92-f9ch-qjcm? 2 open source packages ship an affected version, including p3-symmetric. | Package | Ecosystem | Affected | Fixed in | | --- | --- | --- | --- | | `p3-symmetric` | crates.io | All versions to 0.5.2 | None | | `p3-symmetric` | crates.io | All versions to 0.5.2 | None | ## How do I fix GHSA-3g92-f9ch-qjcm? No fix has been published for GHSA-3g92-f9ch-qjcm yet. Limit exposure of the affected software until one is available. ## Is GHSA-3g92-f9ch-qjcm being exploited? There is no known exploitation of GHSA-3g92-f9ch-qjcm in the wild and no public exploit code. ## Fix prompt for AI coding agents ```text Check this codebase for GHSA-3g92-f9ch-qjcm and fix it if it is affected. What it is: GHSA-3g92-f9ch-qjcm is a vulnerability in p3-symmetric, published on 16 Apr 2026. An attacker can exploit it over the internet without logging in, with no action from the victim. Exploitation: There is no known exploitation of GHSA-3g92-f9ch-qjcm in the wild and no public exploit code. Affected packages and fixed versions: - p3-symmetric (crates.io): affected all versions up to 0.5.2; no fix listed - p3-symmetric (crates.io): affected all versions up to 0.5.2; no fix listed Task: 1. Search the dependency manifests and lockfiles (for example package.json, package-lock.json, requirements.txt, poetry.lock, uv.lock, pom.xml, build.gradle, go.mod, Cargo.lock, Gemfile.lock) and any Dockerfiles for the affected packages, including transitive dependencies. 2. If an affected version is in use, upgrade it to the fixed version for its release line with the project's package manager, so the lockfile is regenerated. Prefer the smallest upgrade that fixes it. 3. If there is no fix, or the upgrade would break the build, do not guess: explain the risk and the safest workaround instead. 4. Run the tests and summarise what changed and why. 5. If nothing in the project is affected, say so and change nothing. Reference: https://insights.sbomify.com/GHSA-3g92-f9ch-qjcm/ (Markdown: https://insights.sbomify.com/GHSA-3g92-f9ch-qjcm/index.md) ``` ## Sources - OSV: https://osv.dev/vulnerability/GHSA-3g92-f9ch-qjcm --- From sbomify Insights: https://insights.sbomify.com/GHSA-3g92-f9ch-qjcm/ Summaries are generated and may contain mistakes; the official description above is the record.