SiYuan CVE-2026-72812

Exploitation requires network access to a SiYuan instance where the API is reachable without prior authentication - specifically, instances exposed on a public or local network interface without an authentication proxy or VPN in front of them. The vulnerable endpoint /api/ref/refreshBacklink must be reachable (default in standard SiYuan deployments). No special account, session token, or application feature needs to be enabled; the flaw exists in the default unauthenticated API surface. Exploitation is limited for instances bound exclusively to localhost or protected by a network-layer authentication gateway, as the attacker cannot reach the endpoint. The impact is constrained to integrity (write operations, backlink corruption) and availability (resource amplification); there is no confidentiality exposure and no code execution primitive identified.

The CVSS 4.0 score of 6.9 (AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:L/VA:L) reflects a network-accessible, zero-complexity attack requiring no privileges or user interaction, which correctly elevates concern despite the Limited integrity and availability impact ratings. Real-world risk is moderate: the direct per-request impact is low (no confidentiality breach, no remote code execution), but the unauthenticated write path and resource amplification potential mean repeated invocations can degrade service stability or corrupt backlink state - a meaningful operational concern for any team relying on SiYuan for knowledge management. No CISA KEV listing and no known public exploit code temper urgency somewhat. EPSS data is not provided in the source intelligence, so exploitation probability cannot be quantified. The key risk multiplier is deployment exposure: SiYuan instances exposed directly to the internet without authentication middleware are fully exploitable by any unauthenticated remote actor; instances bound to localhost or behind a VPN have negligible practical exposure.

An unauthenticated remote attacker identifies a publicly accessible SiYuan instance and sends a crafted HTTP POST request to /api/ref/refreshBacklink with an attacker-controlled block ID. The server processes the request without any authorization check, flushing transaction queues and triggering a global reference scan across all notes. By automating repeated requests with varying block IDs, the attacker induces sustained database write load, degrading application performance or destabilizing backlink state for legitimate users. No public proof-of-concept code has been identified, but the attack requires only standard HTTP tooling.