Skip to main content

CVE-2024-2383: ZenML Clickjacking Vulnerability

CVE-2024-2383 is a clickjacking vulnerability in ZenML that allows attackers to embed the application UI in malicious iframes, potentially tricking users into unauthorized actions. This post explains its impact, affected versions, and mitigation steps.

Published:

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

CVE-2024-2383 Overview

CVE-2024-2383 is a clickjacking vulnerability affecting zenml-io/zenml versions up to and including 0.55.5. The ZenML server fails to set the X-Frame-Options or Content-Security-Policy HTTP response headers. Attackers can embed the ZenML web interface inside an iframe on an attacker-controlled page. Victims who interact with the overlaid page unknowingly trigger actions within their authenticated ZenML session. The issue is categorized under [CWE-1021] (Improper Restriction of Rendered UI Layers or Frames). The ZenML maintainers addressed the issue in version 0.56.3.

Critical Impact

Authenticated ZenML users can be tricked into executing unauthorized actions in the ZenML UI by interacting with an iframe-embedded instance on a malicious page.

Affected Products

  • ZenML (zenml-io/zenml) versions up to and including 0.55.5
  • ZenML Server component (zenml.zen_server.zen_server_api)
  • All deployments exposing the ZenML dashboard over HTTP(S) without upstream frame-protection headers

Discovery Timeline

  • 2024-06-06 - CVE-2024-2383 published to NVD
  • 2026-06-17 - Last updated in NVD database

Technical Details for CVE-2024-2383

Vulnerability Analysis

The ZenML server returns HTTP responses that omit anti-framing controls. Modern browsers rely on either X-Frame-Options: DENY|SAMEORIGIN or a Content-Security-Policy directive containing frame-ancestors to prevent third-party sites from embedding an application in <iframe>, <frame>, or <object> elements. Without those headers, any origin can render the ZenML dashboard inside its own page. An attacker overlays transparent UI elements on top of the embedded ZenML interface and lures an authenticated victim into clicking. The clicks pass through to the ZenML application and execute privileged operations such as pipeline mutations, stack reconfiguration, or user and secret management actions.

Root Cause

The ZenML server built on uvicorn and FastAPI did not inject frame-protection or security response headers. The deployment also exposed the underlying server banner. The combined lack of hardening left the UI eligible for rendering inside hostile frames.

Attack Vector

Exploitation is network-based and requires user interaction. The attacker hosts a page that frames the ZenML dashboard, styles the frame with low opacity, and places decoy controls above sensitive buttons. A logged-in ZenML user who visits the attacker page performs state-changing actions in their own ZenML session without visual indication.

text
# Patch excerpt - docker/base.Dockerfile
-ENTRYPOINT ["uvicorn", "zenml.zen_server.zen_server_api:app",  "--log-level", "debug"]
+ENTRYPOINT ["uvicorn", "zenml.zen_server.zen_server_api:app", "--log-level", "debug", "--no-server-header"]
 CMD ["--port", "8080", "--host",  "0.0.0.0"]

# Patch excerpt - docker/zenml-server-dev.Dockerfile
-ENTRYPOINT ["uvicorn", "zenml.zen_server.zen_server_api:app",  "--log-level", "debug"]
+ENTRYPOINT ["uvicorn", "zenml.zen_server.zen_server_api:app", "--log-level", "debug", "--no-server-header"]
 CMD ["--port", "8080", "--host",  "0.0.0.0"]

Source: ZenML GitHub commit f863fde. The commit adds the --no-server-header flag to the uvicorn entrypoint and introduces security headers in the ZenML server, mitigating UI framing by returning appropriate frame-protection directives.

Detection Methods for CVE-2024-2383

Indicators of Compromise

  • Outbound referrers to the ZenML dashboard originating from unexpected third-party domains in web proxy or CDN logs.
  • HTTP responses from the ZenML server lacking X-Frame-Options and Content-Security-Policy: frame-ancestors headers.
  • Unexpected API state changes (pipeline runs, stack edits, secret rotations) performed by authenticated users shortly after visits to untrusted sites.

Detection Strategies

  • Scan ZenML HTTP responses with a header auditing tool and alert when X-Frame-Options or frame-ancestors directives are absent on any route.
  • Correlate authenticated ZenML session activity with browser referrer data to flag actions originating from embedded contexts.
  • Monitor ZenML audit logs for sequences of state-changing API calls executed in rapid succession with the same session token from a single IP.

Monitoring Recommendations

  • Ingest ZenML server access logs and application audit events into a centralized logging platform for continuous review.
  • Alert on any ZenML server version string reported as 0.55.5 or earlier in asset inventory scans.
  • Track browser security telemetry for users who administer ZenML to detect sites that attempt to frame internal applications.

How to Mitigate CVE-2024-2383

Immediate Actions Required

  • Upgrade all ZenML server deployments to version 0.56.3 or later.
  • Restrict access to the ZenML dashboard to trusted networks, VPNs, or identity-aware proxies until the patch is applied.
  • Enforce short session lifetimes and require re-authentication for sensitive operations in ZenML.

Patch Information

The fix is included in ZenML 0.56.3 and implemented in commit f863fde1269bc355951f8cfc826c0244d88ad5e9. The patch adds security headers to the ZenML server and launches uvicorn with --no-server-header to suppress banner disclosure. Additional context is available on the Huntr bounty report.

Workarounds

  • Place the ZenML server behind a reverse proxy that injects X-Frame-Options: DENY and Content-Security-Policy: frame-ancestors 'none' on every response.
  • Instruct ZenML administrators to use a dedicated browser profile for the dashboard and avoid browsing untrusted sites in the same session.
  • Enable browser-level site isolation and strict SameSite cookie policies to reduce the impact of cross-origin framing.
bash
# Example NGINX reverse proxy snippet enforcing anti-clickjacking headers
server {
    listen 443 ssl;
    server_name zenml.internal.example.com;

    add_header X-Frame-Options "DENY" always;
    add_header Content-Security-Policy "frame-ancestors 'none'" always;
    add_header X-Content-Type-Options "nosniff" always;
    add_header Referrer-Policy "strict-origin-when-cross-origin" always;

    location / {
        proxy_pass http://127.0.0.1:8080;
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
    }
}

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

Experience the Most Advanced Cybersecurity Platform

See how the world’s most intelligent, autonomous cybersecurity platform can protect your organization today and into the future.