GHSA-66R2-5GWJ-GXM2
Vulnerability from github – Published: 2026-09-04 20:50 – Updated: 2026-09-04 20:50A PERMISSIONS ... WHERE clause is evaluated with permission enforcement disabled, so it can't recurse into its own checks. But the clause could also contain data-modifying statements, and these ran with enforcement still off — so evaluating a permission check could write to tables the caller cannot write.
For example:
DEFINE TABLE post PERMISSIONS FOR update
WHERE (CREATE log SET at = time::now()) OR true;
Any user allowed to update a post now also creates a log record, even with no permission on log. The clause is evaluated once per matched record, so one statement can cause several writes.
Impact
Only databases with a PERMISSIONS clause that contains a write are affected; FULL, NONE, and read-only clauses are not.
What an attacker can do:
- With permission to perform the guarded operation (a low-privileged or record user is enough), write to tables in their own database that their permissions would otherwise forbid, by triggering an operation the clause guards.
- Cause several writes from a single statement — the clause is evaluated once per matched record.
- Trigger unintended events, cascades, or data corruption on those tables.
What it can't do:
- Escape the caller's own namespace and database — a permission clause cannot switch namespace or database.
- Perform root- or namespace-level actions such as creating users; the caller's role still applies.
- Read hidden data — this is an integrity issue, not disclosure.
Patches
Permission clauses must now be read-only: defining or importing one that contains a write is rejected, and any write attempted while a clause is evaluated is blocked at runtime, including writes reached through a called function. Read-only clauses are unaffected.
- Versions 3.2.0 and later are not affected by this issue.
Workarounds
Users unable to patch should consider the following workarounds:
- Review your
PERMISSIONSclauses and remove any containingCREATE,UPDATE,DELETE,RELATE,INSERT, orUPSERT. - Limit who can define schema and vet imported data — such a clause must be defined before it can be triggered.
Resources
- DEFINE TABLE … PERMISSIONS
- DEFINE FIELD
fix(core): reject writes in PERMISSIONS clauses and block writes during permission evaluation(included in SurrealDB 3.2.0)
Acknowledgements
Thank you to sondt99 for reporting this issue.
{
"affected": [
{
"package": {
"ecosystem": "crates.io",
"name": "surrealdb-core"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "3.2.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-63733"
],
"database_specific": {
"cwe_ids": [
"CWE-863"
],
"github_reviewed": true,
"github_reviewed_at": "2026-09-04T20:50:40Z",
"nvd_published_at": null,
"severity": "MODERATE"
},
"details": "A `PERMISSIONS ... WHERE` clause is evaluated with permission enforcement disabled, so it can\u0027t recurse into its own checks. But the clause could also contain data-modifying statements, and these ran with enforcement still off \u2014 so evaluating a permission check could write to tables the caller cannot write.\n\nFor example:\n\n```surql\nDEFINE TABLE post PERMISSIONS FOR update\n WHERE (CREATE log SET at = time::now()) OR true;\n```\n\nAny user allowed to update a `post` now also creates a `log` record, even with no permission on `log`. The clause is evaluated once per matched record, so one statement can cause several writes.\n\n### Impact\n\nOnly databases with a `PERMISSIONS` clause that contains a write are affected; `FULL`, `NONE`, and read-only clauses are not.\n\nWhat an attacker **can** do:\n\n- With permission to perform the guarded operation (a low-privileged or record user is enough), write to tables in their own database that their permissions would otherwise forbid, by triggering an operation the clause guards.\n- Cause several writes from a single statement \u2014 the clause is evaluated once per matched record.\n- Trigger unintended events, cascades, or data corruption on those tables.\n\nWhat it **can\u0027t** do:\n\n- Escape the caller\u0027s own namespace and database \u2014 a permission clause cannot switch namespace or database.\n- Perform root- or namespace-level actions such as creating users; the caller\u0027s role still applies.\n- Read hidden data \u2014 this is an integrity issue, not disclosure.\n\n### Patches\n\nPermission clauses must now be read-only: defining or importing one that contains a write is rejected, and any write attempted while a clause is evaluated is blocked at runtime, including writes reached through a called function. Read-only clauses are unaffected.\n\n- Versions 3.2.0 and later are not affected by this issue.\n\n### Workarounds\n\nUsers unable to patch should consider the following workarounds:\n\n- Review your `PERMISSIONS` clauses and remove any containing `CREATE`, `UPDATE`, `DELETE`, `RELATE`, `INSERT`, or `UPSERT`.\n- Limit who can define schema and vet imported data \u2014 such a clause must be defined before it can be triggered.\n\n### Resources\n\n- [DEFINE TABLE \u2026 PERMISSIONS](https://surrealdb.com/docs/surrealql/statements/define/table)\n- [DEFINE FIELD](https://surrealdb.com/docs/surrealql/statements/define/field)\n- `fix(core): reject writes in PERMISSIONS clauses and block writes during permission evaluation` (included in SurrealDB 3.2.0)\n\n### Acknowledgements\n\nThank you to [sondt99](https://github.com/sondt99) for reporting this issue.",
"id": "GHSA-66r2-5gwj-gxm2",
"modified": "2026-09-04T20:50:40Z",
"published": "2026-09-04T20:50:40Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/surrealdb/surrealdb/security/advisories/GHSA-66r2-5gwj-gxm2"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-63733"
},
{
"type": "WEB",
"url": "https://github.com/surrealdb/surrealdb/commit/1e4c3d743e1591f14f340cb627e56d98b6bd7fd7"
},
{
"type": "WEB",
"url": "https://github.com/surrealdb/surrealdb/commit/afea699dfb3c8f274ab36861f8a95f5e98d82f1b"
},
{
"type": "PACKAGE",
"url": "https://github.com/surrealdb/surrealdb"
},
{
"type": "WEB",
"url": "https://github.com/surrealdb/surrealdb/releases/tag/v3.2.0"
},
{
"type": "WEB",
"url": "https://www.vulncheck.com/advisories/surrealdb-before-permissions-bypass-via-permissions-clause"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:L/A:N",
"type": "CVSS_V3"
}
],
"summary": "SurrealDB: Writes in a PERMISSIONS clause bypass table permissions"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
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.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
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.