SiYuan EUVD-2026-60083

Exploitation requires an authenticated account on the target SiYuan instance with permission to upload assets (PR:L), the ability to upload a file containing HTML/JavaScript through the asset-serving endpoint, and - critically - the workspace owner subsequently opening/clicking the malicious asset link (UI:P). The vulnerability is enabled by the server omitting Content-Disposition and X-Content-Type-Options headers when serving assets, so it triggers only when the asset is opened in a browser context that renders inline. Limiting factors: it is not unauthenticated (an attacker needs valid credentials/upload rights), it depends on victim interaction rather than firing automatically, and its practical value is minimal on a single-user instance where uploader and victim are the same person - the real risk is on shared/multi-user deployments prior to 3.7.4.

This is a genuinely high-severity issue rather than a paper-tiger high score. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:P) shows network reachability and low attack complexity, but two real gates temper it: the attacker must already be an authenticated low-privilege user (PR:L), and a victim - the workspace owner - must open the malicious asset link (UI:P). The payoff if those conditions are met is severe: high confidentiality, integrity and availability impact on both the vulnerable and subsequent systems (VC/VI/VA and SC/SI/SA all High), reflecting that script gains full kernel API access. Signals are largely consistent and point to treating this as a real priority for anyone running a multi-user or shared SiYuan instance; a single-user local install where the only account holder is also the only uploader has a much smaller practical exposure, since the attacker and victim would be the same person. No EPSS score, KEV listing, or confirmed public POC is present in the provided data, so exploitation is plausible and low-complexity but not evidenced as active - prioritize patching promptly, but this is not (on current data) a confirmed in-the-wild campaign.

A low-privileged member of a shared SiYuan workspace uploads a crafted .html file (or a file whose content is HTML) as an asset and shares or embeds its link. When the workspace owner clicks that asset link, the browser renders it inline and the embedded JavaScript executes in SiYuan's origin, invoking kernel APIs to read notes, exfiltrate data, or alter the workspace. Attack complexity is low, but success depends on the victim opening the link; no confirmed public POC is present in the provided data.