GHSA-W437-9MWM-J6VQ
Vulnerability from github – Published: 2026-11-12 19:42 – Updated: 2026-11-12 19:42Impact
Documents belonging to one caller could be returned to another. Where an application passes a customer or tenant identifier as a search filter, that scoping could be defeated by ordinary concurrent traffic. No injection and no malformed input are required — normal load is enough.
The BM25 keyword index was a single per-pipeline object shared by every request. A request cleared it, refilled it from its own filtered corpus and then searched it, taking a separate lock for each step rather than holding the sequence as a unit. Concurrent requests therefore interleaved, and a search could run against a corpus that a different request's filter had populated.
Who is affected: any deployment of pgEdge RAG Server at 1.0.0 or earlier, including all alpha and beta pre-releases, that serves concurrent queries — most acutely those relying on a search filter to separate tenants or customers. The query endpoint is unauthenticated in the affected versions (internal/server/middleware.go:37 applies logging, panic recovery and optional CORS only), so both the party whose results leak and the party receiving them are anonymous remote callers.
Patches
Fixed in 2.0.0. Upgrade to 2.0.0 or later.
The index is now built per request rather than shared. The regression is pinned by TestSearch_ConcurrentRequestsDoNotShareKeywordIndex in internal/pipeline/orchestrator_test.go.
Upgrading does not tell a developer whether or not their application affected, and needs action beyond upgrading. If a consuminag application scopes results by tenant, treat responses served under concurrent load on any affected version as potentially having included another tenant's documents. Affected responses cannot be identified from the server's own logs, so there is no retrospective check available.
Workarounds
Set hybrid_enabled: false on the pipeline, which skips the keyword arm and with it the shared index. Callers can also pass disable_hybrid per request, which can only ever turn the arm off.
Neither is a substitute for upgrading.
{
"affected": [
{
"package": {
"ecosystem": "Go",
"name": "github.com/pgEdge/pgedge-rag-server"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "1.0.1-0.20260729020411-a09d4872356b"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-125325"
],
"database_specific": {
"cwe_ids": [
"CWE-362",
"CWE-488"
],
"github_reviewed": true,
"github_reviewed_at": "2026-11-12T19:42:07Z",
"nvd_published_at": "2026-11-10T18:17:33Z",
"severity": "MODERATE"
},
"details": "### Impact\n\nDocuments belonging to one caller could be returned to another. Where an application passes a customer or tenant identifier as a search filter, that scoping could be defeated by ordinary concurrent traffic. **No injection and no malformed input are required** \u2014 normal load is enough.\n\nThe BM25 keyword index was a single per-pipeline object shared by every request. A request cleared it, refilled it from its own filtered corpus and then searched it, taking a separate lock for each step rather than holding the sequence as a unit. Concurrent requests therefore interleaved, and a search could run against a corpus that a different request\u0027s filter had populated.\n\n**Who is affected:** any deployment of pgEdge RAG Server at `1.0.0` or earlier, including all alpha and beta pre-releases, that serves concurrent queries \u2014 most acutely those relying on a search filter to separate tenants or customers. The query endpoint is unauthenticated in the affected versions (`internal/server/middleware.go:37` applies logging, panic recovery and optional CORS only), so both the party whose results leak and the party receiving them are anonymous remote callers.\n\n### Patches\n\n**Fixed in `2.0.0`. Upgrade to `2.0.0` or later.**\n\nThe index is now built per request rather than shared. The regression is pinned by `TestSearch_ConcurrentRequestsDoNotShareKeywordIndex` in `internal/pipeline/orchestrator_test.go`.\n\n**Upgrading does not tell a developer whether or not their application affected, and needs action beyond upgrading.** If a consuminag application scopes results by tenant, treat responses served under concurrent load on any affected version as potentially having included another tenant\u0027s documents. **Affected responses cannot be identified from the server\u0027s own logs**, so there is no retrospective check available.\n\n### Workarounds\n\nSet `hybrid_enabled: false` on the pipeline, which skips the keyword arm and with it the shared index. Callers can also pass `disable_hybrid` per request, which can only ever turn the arm off.\n\nNeither is a substitute for upgrading.",
"id": "GHSA-w437-9mwm-j6vq",
"modified": "2026-11-12T19:42:08Z",
"published": "2026-11-12T19:42:07Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/pgEdge/pgedge-rag-server/security/advisories/GHSA-w437-9mwm-j6vq"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-125325"
},
{
"type": "WEB",
"url": "https://github.com/pgEdge/pgedge-rag-server/pull/44"
},
{
"type": "WEB",
"url": "https://github.com/pgEdge/pgedge-rag-server/commit/a09d4872356b510c81f11591845445c33ff45a84"
},
{
"type": "PACKAGE",
"url": "https://github.com/pgEdge/pgedge-rag-server"
},
{
"type": "WEB",
"url": "https://github.com/pgEdge/pgedge-rag-server/releases/tag/v2.0.0"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N",
"type": "CVSS_V3"
}
],
"summary": "pgEdge RAG Server shared keyword index leaked results between concurrent unauthenticated requests"
}
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.