SiYuan CVE-2026-84803

Exploitation requires an authenticated user account with file upload privileges in the SiYuan instance (PR:L per CVSS vector). A second user (the victim) must open or preview the uploaded asset in a browser session where they are logged into SiYuan (UI:P per CVSS vector). The vulnerability is conditional on the target file extension being one of the bypassing types (.xht, .ehtml, .xsl, .xbl, .rdf) and the browser's MIME-type mapping rendering it as an executable document. Single-user self-hosted deployments where no other party has upload access face no practical risk. Multi-tenant or team deployments with untrusted upload-capable users are the realistic attack surface.

The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H) reflects low-complexity network exploitation requiring only low-privilege authentication and passive victim interaction, with high confidentiality, integrity, and availability impact on the vulnerable system. Real-world risk is most significant in multi-user or team SiYuan deployments where untrusted collaborators can upload files - single-user self-hosted instances face negligible risk since the attacker and victim would be the same person. No EPSS score is present in the provided data; no KEV listing exists; no public exploit code has been confirmed. The bypass technique (uploading unlisted executable-capable extensions) is technically straightforward for anyone familiar with browser MIME-type handling, making the attack low-effort once access is obtained. The PR:L prerequisite is the primary mitigating factor for mass exploitation scenarios.

An authenticated collaborator in a shared SiYuan workspace uploads a crafted file named payload.xsl containing an embedded JavaScript payload that reads document.cookie and posts the API token to an attacker-controlled endpoint. When a workspace admin or co-user opens the shared asset link in their browser, the browser renders the .xsl file as an executable XML document and runs the script, silently exfiltrating the victim's API token. No public exploit code has been confirmed, but the technique is straightforward given the described extension bypass.