OSV 1.4.0 · github-reviewed · 修改于 2026-09-04 07:04
发布时间
2026-09-04 07:04
GitHub 审查时间
2026-09-04 07:04
NVD 发布时间
—
源文件
advisories/github-reviewed/2026/09/GHSA-5fhr-f75j-8wr9/GHSA-5fhr-f75j-8wr9.json
CVE: This vulnerability corresponds to CVE-2026-72800.
POST /api/av/getAttributeViewKeysByID returns a database's full column schema with no publish-access filtering, while its sibling getAttributeViewKeys applies the filter for reader sessions. Two further endpoints, getBlockDefIDsByRefText and getBlockRelevantIDs return workspace-wide block IDs with no publish scoping. All three are CheckAuth-only, so they are reachable by the publish RoleReader token and by the anonymous account when Publish.Auth.Enable is false.
(a) getAttributeViewKeysByID: ungated column-schema disclosure
The entire handler (kernel/api/av.go, router line 548):
ret.Data = model.GetAttributeViewKeysByID(avID, keyIDs...) // no publish gate
With an empty keyIDs, GetAttributeViewKeysByID calls av.ParseAttributeView(avID) and returns every column's *av.Key:
| Field | Discloses |
|---|---|
Name, Desc | column title and user-authored description |
Options | the full single/multi-select vocabulary and colours (e.g. status labels such as "Fired", "Confidential") |
Template | the column's Sprig template expression — logic and field references |
Relation | the target avID, allowing pivot to another database |
Rollup, NumberFormat, date config | further schema |
Guarded-sibling asymmetry, same file: getAttributeViewKeys (router line 526) runs when . applies nothing. A reader who knows an — trivially harvested from the attribute of any published document embedding a database obtains the schema of any database in the workspace, including those whose host documents are hidden, password-protected, or publish-forbidden. targets allow walking to sibling databases.
FilterBlockAttributeViewKeysByPublishAccess(...)IsReadOnlyRoleContextgetAttributeViewKeysByIDavIDdata-av-idRelation(b) Two unscoped block-ID enumeration oracles
Both CheckAuth-only with no publish gate. Neither returns content of its own, but both yield valid block IDs that other endpoints turn into content:
getBlockDefIDsByRefText (router line 237) → GetBlockDefIDsByRefText(anchor) returns the block IDs, workspace-wide and including private documents, whose reference/anchor text equals a caller-supplied string: a ref-text → ID oracle.getBlockRelevantIDs (router line 274) → GetBlockRelevantIDsInBox(id, <notebook from request>) returns parent/previous/next block IDs and traverses decrypted encrypted-notebook structure while the notebook is unlocked, a tree-walk oracle.Verified at origin/master (eef105683): all three handler bodies contain no publish-access, publish-ignore, or readonly-role check; all three routes are registered CheckAuth without CheckAdminRole.
Precondition: publish mode enabled (default port 6808); anonymous when Publish.Auth.Enable is false, otherwise any publish reader account. A database exists whose host document is publish-forbidden or password-protected.
(a) Column schema of any database:
POST http://127.0.0.1:6808/api/av/getAttributeViewKeysByID
{"avID":"<AV_ID>"}
Returns every column's key object names, descriptions, select vocabularies, template expressions, and Relation target avIDs for a database whose host document the reader may not access.
Control: the sibling getAttributeViewKeys with the same avID returns filtered results for the same reader session, confirming the boundary is enforced there and omitted here.
(b) ID oracles:
POST http://127.0.0.1:6808/api/block/getBlockDefIDsByRefText
{"anchor":"<known ref text>"}
→ block IDs workspace-wide, including blocks in private documents
POST http://127.0.0.1:6808/api/block/getBlockRelevantIDs
{"id":"<BLOCK_ID>","notebook":"<BOX_ID>"}
→ parent/previous/next block IDs; traverses encrypted-notebook structure when unlocked
An anonymous reader (publish mode with auth disabled) or any publish RoleReader can read the complete column schema of any database in the workspace, including column descriptions, select-option vocabularies (which frequently encode sensitive category labels), template logic, and relation targets that permit pivoting to further databases regardless of whether the host document is hidden, password-protected, or excluded from publishing.
The two enumeration endpoints additionally supply valid block IDs across the publish boundary, including from encrypted notebooks while unlocked. This removes the "attacker must already know a valid ID" precondition from other block-read endpoints, converting ID knowledge into content disclosure. Confidentiality-only.
getAttributeViewKeysByID is a distinct handler from getAttributeView; although column definitions also appear within the latter's response payload, a fix applied to one handler does not remediate the other, and getAttributeViewKeysByID has its own filtered sibling (getAttributeViewKeys) demonstrating the intended treatment.
getAttributeViewKeysByID with FilterBlockAttributeViewKeysByPublishAccess, mirroring getAttributeViewKeys.getBlockDefIDsByRefText and getBlockRelevantIDs to publish-accessible blocks for reader sessions, and ensure the *InBox traversal path applies the same check before walking encrypted-notebook structure.