SiYuan CVE-2026-82654

Exploitation requires the attacker to have a low-privileged authenticated account on the SiYuan instance (PR:L) with the ability to create or edit a block and set its name, alias, or memo field. Execution is not automatic: a second, different user must view or open a document that references or displays the poisoned block through the hint, backlink, or breadcrumb rendering functions (UI:P/UI:R), so a single-user local install is effectively not exploitable. Limiting factors: the need for a second victim user, the requirement that the victim navigate to a referencing view, the authenticated-account prerequisite, and affected versions being strictly before v3.8.1.

The vendor CVSS 4.0 score of 9.3 (AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:L, with subsequent-system SC:H/SI:H) is high, but the raw number should be read against real conditions: exploitation requires an authenticated account (PR:L) and victim interaction (UI:P) - namely a second user opening or viewing a document that surfaces the malicious block. This is therefore most impactful in multi-user or shared/collaborative SiYuan deployments; a purely single-user local install has little exposure since the attacker and victim would be the same person. No EPSS score, KEV entry, or POC was provided, so predicted exploitation likelihood cannot be quantified from the input and there is no evidence of active exploitation. The high subsequent-system impact scoring reflects that script execution in the victim's authenticated context can lead to account/session compromise. Overall this is a genuine, patch-now issue for any shared SiYuan instance, but the account-plus-interaction prerequisites make it lower urgency for isolated single-user setups than the 9.3 alone would imply.

An attacker with a low-privileged account on a shared SiYuan instance creates a block and sets its name (or alias/memo) to a string containing script tags. When a colleague later opens or edits a document that references that block - triggering the backlink, breadcrumb, or hint rendering - the script executes in the colleague's authenticated session, letting the attacker steal session data or perform actions as that user. No public POC was provided, but the technique is straightforward given the described lack of output escaping.