Description

SiYuan before v3.7.4 fails to set Content-Disposition and X-Content-Type-Options headers when serving arbitrary file assets, allowing stored cross-site scripting attacks. Authenticated attackers can upload HTML files as assets and execute scripts with full kernel API access when the workspace owner opens the asset link.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "9.0",
  "pubDate": "2026-08-17T11:16:40.187Z",
  "pubdate": "2026-08-17T11:16:40.187Z",
  "executiveSummary": "SiYuan before v3.7.4 suffers from a stored cross-site scripting (XSS) vulnerability arising from the improper handling of arbitrary file asset serving. Specifically, the application fails to enforce security-critical HTTP response headers, namely Content-Disposition and X-Content-Type-Options, when transmitting uploaded asset files to clients.\nThis security oversight allows authenticated attackers to upload malicious HTML files disguised or handled as benign assets within the application. When a workspace owner accesses the asset link, the browser renders the HTML context directly, executing arbitrary JavaScript payloads in the security context of the victim's session.\nThe impact of successful exploitation is critical, as the executing script gains full kernel API access within the application environment. This grants the attacker comprehensive control over the victim's workspace and underlying system interactions, posing significant confidentiality, integrity, and availability risks.\nExploitation requires authentication to the platform to perform the initial file upload, followed by user interaction where the workspace owner must be induced to open the malicious asset link. Organizations utilizing affected instances face severe operational risk due to the breadth of privileges accessible via the application API upon successful payload execution.",
  "technicalDetails": "The vulnerability resides in the asset-serving mechanism of SiYuan prior to version v3.7.4, which processes and delivers user-uploaded files without enforcing robust MIME-type sniffing protections or forcing safe content delivery behaviors.\nRoot Cause: The lack of the X-Content-Type-Options response header (typically set to nosniff) allows web browsers to perform MIME-type sniffing on served assets, potentially interpreting non-executable or generic data types as active content such as text/html. Furthermore, the absence of a restrictive Content-Disposition header prevents the browser from forcing a download prompt for potentially dangerous file types, causing inline rendering of untrusted markup.\nAttack Flow: 1. An authenticated attacker leverages standard upload functionalities to store a malicious HTML file containing crafted JavaScript payloads within the SiYuan asset repository. 2. The attacker induces the workspace owner to click or navigate to the specific URL pointing to the uploaded asset. 3. The SiYuan server delivers the asset file to the victim's browser without setting X-Content-Type-Options or Content-Disposition: attachment. 4. The victim's browser interprets the file as HTML and executes the embedded JavaScript payload within the origin context of the application.\nPrivilege Requirements and Exposure: Exploitation requires an authenticated attacker capable of uploading files to the workspace. Although authentication is required, standard users or compromised low-privilege accounts can execute the attack vector. The network exposure encompasses any deployment where users interact with uploaded workspace assets.\nPayload Behavior and Impact: Upon execution within the victim browser session, the malicious script leverages full kernel API access exposed by the application. This post-exploitation capability allows the attacker to interact with internal system primitives, manipulate workspace data, exfiltrate sensitive information, or perform actions with the administrative privileges of the workspace owner."
}