CVE-2026-72790

The endpoint `/api/notebook/getNotebookInfo` exposes metadata about closed (private) notebooks without authentication. This includes notebook names, document counts, storage sizes, and creation/modification timestamps. If your SiYuan instance has this endpoint reachable over the network, an unauthenticated attacker can enumerate all notebooks on the server regardless of their closed status. The CVSS 5.8 score is misleading for this vulnerability class. In a knowledge management tool, notebook metadata functions as partial content disclosure rather than generic system information. A notebook named 'Sources - Investigation Alpha' with 47 documents signals exactly what someone was working on and at what scale — that's reconnaissance-gold for anyone targeting researchers, investigators, or journalists. The EPSS score of 0.00237 compounds this misunderstanding: it reflects generic web application exploitability, but SiYuan is a niche tool with a high-value userbase. The relevant threat model isn't random internet scanning — it's targeted reconnaissance against someone already identified as a SiYuan user. This vulnerability likely emerged from architectural drift. The endpoint probably predates SiYuan's 'closed notebook' privacy feature — it was correct behavior when notebooks were inherently private, and no one audited existing API endpoints when the privacy controls were added. Check whether other notebook-related endpoints (`/api/document/*`, `/api/asset/*`) respect closed notebook controls. If they don't, you're looking at a pattern failure, not a single oversight. Prioritize remediation based on deployment context. Self-hosted instances accessible to the internet carry the highest risk. If your threat model includes targeted adversaries (investigative journalists, researchers working on sensitive topics, anyone storing work product they wouldn't want enumerated), treat this as higher severity than the base score indicates. Consider whether users who created closed notebooks under the assumption of protection should be notified, since they may have stored work they believed was private while this endpoint was exposed.

Reviewed through automated stages and approved by a human before publication.