/

CVE-2026-72804

Severity:

HIGH RISK

(8.6)

Information Disclosure (CWE-200) in SiYuan Graph APIs

CVE-2026-72804

Deferred

Vulnerability type

Category:

Information Disclosure

CWE:

CWE-200

Impact:

Unauthorized Disclosure of Protected Document Content

Severity:

HIGH RISK

(8.6)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

SiYuan versions before v3.7.4 do not consistently validate the publish-password tier for the getGraph and getLocalGraph endpoints. An unauthenticated reader may call these endpoints to retrieve block-level content and the complete reference topology of password-protected documents.

Affected context

The issue occurs in SiYuan deployments where password-protected documents are published or exposed through the graph-related endpoints. Risk is higher when those endpoints are reachable by anonymous users and the application relies on the publish-password tier to restrict access to document content and reference relationships.

Exploit preconditions

  • Attack surface: Network-accessible getGraph and getLocalGraph endpoints
  • Required access: No authentication or valid document password may be required
  • Trigger point: A request for graph or local-graph data associated with a password-protected document
  • Why exploitation is possible: The endpoints fail to enforce the required publish-password authorization before returning block content and document reference topology

Why this happens

The vulnerability results from incomplete authorization and access-control validation on graph data endpoints. The application treats a request as eligible for processing without verifying that the caller has satisfied the publish-password tier protecting the referenced document, exposing data intended to remain restricted. This behavior maps to CWE-200: Exposure of Sensitive Information to an Unauthorized Actor.

Potential impact

  • Unauthorized readers may access protected block-level document content
  • Attackers could reconstruct relationships among documents and blocks
  • Confidential notes, metadata, or business information may be disclosed
  • Exposed reference topology could support further targeted access attempts
  • Organizations may experience privacy, compliance, or intellectual-property impacts

General mitigation guidance

Mitigation generally focuses on enforcing authorization consistently at every endpoint that returns protected document data.

Common defensive patterns include:

  • Require successful publish-password validation before graph data is queried or serialized
  • Apply centralized, deny-by-default authorization checks to both graph endpoints
  • Ensure authorization covers indirect metadata and reference topology, not only document bodies
  • Minimize returned fields and avoid exposing protected content to anonymous callers
  • Use least-privilege service identities and restrict public exposure of administrative or content APIs
  • Add regression tests for anonymous, incorrect-password, and authorized access paths

How teams detect issues like this

Organizations can assess this risk through:

  • Static and semantic analysis for endpoint handlers that return document data without authorization checks
  • Code review searches for graph or metadata APIs that bypass shared access-control middleware
  • Tests verifying that anonymous and invalid-password requests receive no protected content or topology
  • API gateway and application policies requiring authentication or authorization for sensitive document resources
  • Runtime monitoring for repeated anonymous calls to graph endpoints and unusual enumeration patterns
  • Automated security testing that compares responses across authorization states

Sources

SourceLink
github.comgithub.com/siyuan-note/siyuan/security/advisories/GHSA-vpjw-wf5h-cgpq
vulncheck.comvulncheck.com/advisories/siyuan-before-authentication-bypass-via-graph-endpoints

Frequently asked

questions

Yes, when the affected graph endpoints are reachable over the network. The described behavior permits anonymous requests, so a remote attacker may not need an account.

The attacker generally needs network access to the SiYuan service and knowledge of a target document or endpoint request format. Supplying the publish password is not expected to be necessary.

Enforce the publish-password authorization decision before retrieving or serializing any block content, metadata, or reference topology, using a shared deny-by-default access-control path.

Treat indirect data such as graphs and references as protected content, test every endpoint under anonymous and unauthorized conditions, and enforce authorization policies consistently at the application or gateway layer.

Find out what's already
exploitable in your codebase.