SiYuan CVE-2026-68585

The SiYuan instance's /api/block/getBlockInfo endpoint must be reachable over the network by the attacker - no special network position is required if the instance is internet-exposed. The attacker must supply a valid block ID referencing a publish-excluded document; block IDs are opaque identifiers that may require enumeration. Exploitation is possible with zero credentials: the description explicitly states anonymous readers (no token) or holders of a publish RoleReader token can trigger the disclosure. What limits exploitation: the disclosure is metadata-only (title, path, icon, root ID) and does not expose document body content; the attacker must know or enumerate valid block IDs; instances deployed exclusively on internal networks without internet exposure are not reachable by external anonymous attackers.

The CVSS 4.0 score of 6.9 with vector AV:N/AC:L/AT:N/PR:N/UI:N reflects that this is a trivially accessible, unauthenticated endpoint requiring only a valid block ID. The impact is bounded strictly to metadata - title, path, root ID, icon - not document body content, which limits the blast radius substantially compared to a full content disclosure. No EPSS score was provided in the input, so exploitation probability cannot be quantified precisely. The absence of a KEV listing and no known POC suggests this has not been leveraged in observed attacks. The primary realistic risk is enumeration of sensitive document titles and organizational hierarchy from a deliberately restricted knowledge base, which could facilitate targeted social engineering or reveal confidential project names. This is a genuine authorization control bypass, but real-world impact depends on the sensitivity of the document metadata and the exposure of the SiYuan instance to untrusted networks.

An external attacker with no credentials queries GET /api/block/getBlockInfo?id=<blockID> against a network-exposed SiYuan instance using a block ID obtained through enumeration or prior knowledge. The server returns the full document root metadata - including title, notebook, path, and root ID - for a document the administrator explicitly excluded from publishing. The attacker can repeat this process across multiple block IDs to reconstruct the hidden document structure of the knowledge base.