SiYuan EUVD-2026-52278

Exploitation requires reaching the /api/filetree/searchDocs endpoint, which is possible in one of two ways: (1) fully unauthenticated when the instance runs in publish mode with the configuration setting Publish.Auth.Enable set to false, or (2) with a publish RoleReader token when publish auth is enabled. The keyword parameter is the injection vector. Data-modification impact additionally depends on the target notebooks being cleartext (non-encrypted) - encrypted notebooks are not affected - and on the driver's support for stacked statements over the read-write SQLite handle, both of which are default in affected versions. Limiting factors: an instance not running publish mode, or one with Publish.Auth.Enable=true and no leaked RoleReader token, substantially narrows or removes anonymous exposure.

This is a genuine high priority, not an inflated CVSS number. The vendor CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:N) describes trivial remote exploitation with no attack requirements, and the VC:H/VI:H confidentiality-and-integrity impact matches the described read-and-modify capability over a read-write SQLite handle. The critical nuance is the privileges axis: PR:N holds only in the specific publish-mode configuration where Publish.Auth.Enable is false; otherwise a publish RoleReader token is required, which is a low-privilege but non-zero barrier - so real-world exposure depends heavily on deployment. No EPSS score, KEV listing, or POC was provided, so likelihood signals beyond the CVSS vector are absent. Net: high technical severity and low complexity, but blast radius is gated by whether an instance runs publish mode and how authentication is configured.

An attacker who can reach a SiYuan instance running publish mode with Publish.Auth.Enable=false sends an HTTP request to /api/filetree/searchDocs with a crafted keyword containing SQL metacharacters and a stacked statement. Given the low attack complexity and no required user interaction, they read sensitive note contents and then append a second semicolon-separated statement to modify or delete records across all cleartext notebooks. No public exploit code has been identified, but the injection is straightforward to reproduce from the advisory details.