Siyuan CVE-2026-32938

SiYuan desktop versions ≤3.6.0 with publish-service enabled. Attacker must be authenticated as a publish-service visitor. Victim must paste HTML containing file:// links pointing to sensitive files (e.g., config files, credentials) via the /api/lute/html2BlockDOM endpoint. No user interaction beyond pasting required.

The CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:L/A:H) indicates network-based exploitation with low attack complexity requiring only low privileges and no user interaction, with scope change amplifying impact. The 9.9 critical score reflects the cross-scope confidentiality breach potential. However, real-world risk factors suggest moderate urgency: the attack requires prior authentication (PR:L), limiting exposure to authenticated users or those with publish-service visitor access; the vulnerability affects a desktop application rather than internet-facing infrastructure; and there is no evidence of public proof-of-concept code, EPSS probability data, or KEV listing indicating active exploitation. The primary risk exists in environments where untrusted authenticated users have access to SiYuan instances, particularly in shared or collaborative deployments. Organizations using SiYuan for sensitive knowledge management should prioritize patching, but the lack of exploitation signals suggests this is not an immediate wormable or mass-exploitation threat.

An authenticated attacker with publish-service visitor credentials crafts malicious HTML content containing file:// links pointing to sensitive system files such as /etc/passwd, SSH private keys in ~/.ssh/id_rsa, or application configuration files. When this HTML is pasted and processed by the /api/lute/html2BlockDOM endpoint, the desktop kernel copies these files into the workspace assets directory without path validation. The attacker then retrieves the exfiltrated files by issuing authenticated GET requests to /assets/ followed by the predictable file paths, successfully stealing sensitive data from the victim's system.