Cap's AI Video Metadata Exposure via IDOR

Executive Summary

This vulnerability exists in Cap's GET /api/video/ai endpoint, which fails to verify whether a user owns or is a member authorized to access certain video AI metadata. As a result, authenticated attackers can provide arbitrary video IDs to access private AI-generated content such as titles, summaries, and chapters that they should not see.

Additionally, attackers can trigger unauthorized AI generation processes that consume the video owner's credits without their consent.

Useful 0 Not Useful 0

Impact Analysis

The vulnerability can lead to unauthorized disclosure of sensitive AI-generated video metadata, compromising privacy and confidentiality.

It can also result in financial impact or resource misuse, as attackers can trigger AI generation that consumes the video owner's credits without permission.

Useful 0 Not Useful 0

Compliance Impact

This vulnerability allows authenticated attackers to access private video AI metadata without proper validation of user ownership or membership. This unauthorized access to sensitive AI-generated content could lead to exposure of personal or confidential information.

Such unauthorized disclosure of sensitive data may result in non-compliance with data protection regulations like GDPR and HIPAA, which require strict controls over access to personal and sensitive information.

Additionally, the unauthorized triggering of AI generation consuming the video owner's credits without consent may raise concerns about consent and resource misuse under these regulations.

Useful 0 Not Useful 0

Detection Guidance

This vulnerability can be detected by monitoring requests to the GET /api/video/ai endpoint for unauthorized access attempts. Specifically, look for authenticated users supplying arbitrary video IDs to access AI-generated metadata that they should not have permission to view.

You can detect potential exploitation by checking your server logs or network traffic for unusual or unauthorized API calls to /api/video/ai with video IDs that do not belong to the requesting user.

While no explicit commands are provided in the resources, a general approach would be to use tools like curl or HTTP request logging to query the endpoint with various video IDs and observe if unauthorized data is returned.

• Example curl command to test access: curl -H "Authorization: Bearer <token>" https://<your-cap-instance>/api/video/ai?videoId=<arbitrary_video_id>
• Review server or application logs for repeated or suspicious access patterns to /api/video/ai endpoint.

Useful 0 Not Useful 0

Mitigation Strategies

Immediate mitigation steps include applying the security fix that enforces proper authorization checks on the /api/video/ai endpoint and related transcript endpoints.

This fix involves implementing the `canView` policy to ensure only authorized users can access video AI metadata and transcripts.

Additionally, rate limiting should be applied to the transcript translation endpoint to prevent abuse and unauthorized AI generation that consumes credits.

If you are running a self-hosted Cap instance, update your deployment to include the patch from commit 8d48642b6e7938af238386383ef1c273be4110dd or later, which contains the fix for this vulnerability.

Ensure consistent error handling is in place to avoid leaking information about private or protected videos.

Monitor your system for unauthorized access attempts and consider temporarily restricting access to the affected endpoints until the fix is applied.

Useful 0 Not Useful 0