Stored XSS to RCE in SiYuan attribute-view cell rendering
CVE-2026-54158 is a stored cross-site scripting vulnerability in SiYuan, an open-source personal knowledge management system, affecting versions prior to 3.7.0. The flaw is in the attribute-view database cell renderer function genAVValueHTML(), which interpolates untrusted cell content without proper escaping in at least four rendering branches: text, url, phone, and mAsset. Crafted cell values can break out of surrounding HTML context and execute arbitrary JavaScript when a user opens the block-attribute panel containing the malicious row. The malicious content is also not escaped on input, allowing the payload to persist unchanged in attribute-view data and propagate through normal workspace synchronization. In Electron desktop builds, the renderer executes with nodeIntegration enabled, allowing the XSS condition to be chained into host-level remote code execution through Node.js APIs.
What it means and what to do now: patch path, interim mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
Successful exploitation allows execution of arbitrary JavaScript in the victim's SiYuan renderer context when the affected attribute-view content is viewed. Because the payload is stored in synchronized workspace data, a single malicious entry can persist and trigger across multiple devices that open the relevant block-attribute panel. On Electron desktop deployments, the impact escalates beyond browser-context script execution to host remote code execution due to nodeIntegration being enabled, potentially resulting in full compromise of the local user context, including access to local files, command execution, data theft, tampering, and disruption. On non-Electron web-exposed contexts, the impact is limited to stored XSS in the application context.
If you can’t patch tonight, do this now.
Until patched, restrict write access to shared or synchronized workspaces to trusted users only, and reduce or disable synchronization paths that permit untrusted content injection. Avoid opening block-attribute panels for untrusted rows or workspaces. Review synchronized attribute-view storage for malicious entries and remove any injected payloads. Where feasible, harden Electron deployments by disabling or isolating dangerous renderer capabilities such as Node.js integration, and enforce strict output encoding and URL scheme validation for rendered cell content as compensating controls.
Patch, then assume compromise.
Upgrade SiYuan to version 3.7.0 or later, which fixes the vulnerable rendering behavior. The vulnerable renderer branches in genAVValueHTML() should properly escape untrusted content before inserting it into HTML or attribute contexts, specifically for text, url, phone, and mAsset cell types. Defense-in-depth measures include validating and sanitizing cell content before persistence so malicious payloads are not stored verbatim, but renderer-side output encoding is the primary corrective action because backend-only changes do not neutralize already-synced malicious data.
No public exploits tracked yet. Mallory keeps watching.
5 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.