SiYuan EUVD-2026-57465

Exploitation requires that the SiYuan instance has the publish feature enabled and contains at least one document configured as publish-forbidden or password-protected. The attacker must have network-level access to the SiYuan HTTP API (default port 6806). The CVSS 4.0 PR:N metric confirms no authentication is required - anonymous users with no account can exploit this. The attacker must supply valid block IDs, which may require prior enumeration or guessing, representing the primary practical barrier. Exploitation is bounded in impact: only metadata and content snippets (titles, ancestor block text, reference text, path metadata) are leaked - full document content exfiltration is not confirmed. Deployments that do not use the publish feature, that restrict API access to localhost or a trusted network, or that are running v3.7.4 or later are not at risk.

The CVSS 4.0 base score of 6.9 with vector AV:N/AC:L/AT:N/PR:N/UI:N/VC:L/VI:N/VA:N/SC:L/SI:N/SA:N accurately captures the key risk profile: network-accessible, no authentication, no complexity barrier, with low confidentiality impact to both the vulnerable and subsequent systems. The low confidentiality rating is appropriate - the leaked data is metadata and content snippets rather than full plaintext document bodies, limiting the blast radius of any single exploitation. No EPSS score was included in the provided intelligence, and there is no CISA KEV entry, meaning active exploitation is not confirmed. No public exploit code was referenced, reducing near-term opportunistic risk. Real-world priority is moderate for organizations running network-exposed SiYuan instances with publish controls in use; the missing authorization class (CWE-862) is typically straightforward to exploit once block IDs are known or guessable. Deployments bound to localhost, behind a firewall, or not using the publish feature carry negligible practical risk despite the network-level vector.

An attacker with network access to a SiYuan instance sends unauthenticated HTTP requests to the getBlockBreadcrumb, getRefText, or getBlockTreeInfos API endpoints, supplying block IDs that may be guessed, enumerated, or previously observed. The server returns document titles, breadcrumb path metadata, ancestor block content snippets, and reference text for documents whose owners configured them as publish-forbidden or password-protected. No publicly available exploit code has been identified, but the attack requires only HTTP request crafting and knowledge of valid block IDs.