SiYuan CVE-2026-100636

Requires authenticated administrator access to the SiYuan kernel and use of the exportBrowserHTML endpoint with a folder parameter containing directory-traversal sequences (e.g., ../../). Exploitation is limited to overwriting index.html in locations that already exist and are writable by the kernel process (pre-existing kernel-writable paths), and is gated by administrator authentication (PR:H) - there is no unauthenticated or user-interaction path. Affects SiYuan versions before v3.8.4; upgrading to v3.8.4 remediates.

This is a moderate, config-independent but high-privilege issue. The vendor CVSS 4.0 base score of 8.3 (AV:N/AC:L/AT:N/PR:H/UI:N/VC:N/VI:H/VA:L/SC:N/SI:H/SA:L) reflects network reach and low complexity, but the decisive limiter is PR:H - exploitation requires an already-authenticated SiYuan administrator, which substantially narrows the real-world threat since an admin already holds broad control over the instance. There is no EPSS score, no CISA KEV listing, and no public exploit code identified at time of analysis, so the '8.3' should be read as a capability rating rather than an indicator of imminent, widespread exploitation. No SSVC decision data was supplied. The signals are internally consistent (no confidentiality impact, integrity-dominant, write-primitive), and the main disagreement is that the vendor's SI:H (subsequent-system integrity high) is somewhat generous - the write primitive is powerful but constrained to pre-existing kernel-writable locations and gated behind admin auth. Net: a genuine path-traversal/stored-XSS bug worth patching, but not a drop-everything priority absent exploitation evidence.

An authenticated SiYuan administrator (or an attacker who has hijacked an admin session) calls the exportBrowserHTML endpoint with a folder parameter containing directory-traversal sequences (e.g., ../../), escaping the intended export directory to overwrite an index.html in another kernel-writable path. The planted HTML delivers stored XSS to anyone who later loads that page, or simply defaces the workspace. No public exploit code is currently identified, and the attack requires high privileges.