Improper Access Control in Cognee Before 1.2.0

Executive Summary

This vulnerability exists in Cognee versions before 1.2.0 and is due to improper access control. It allows unauthenticated attackers to overwrite the global Large Language Model (LLM) provider configuration by self-registering an account and calling the settings endpoint, which lacks admin or superuser verification.

By exploiting this, attackers can redirect all LLM operations across the entire instance to an attacker-controlled endpoint.

This happens because the configuration cache is a process-wide singleton, meaning the change affects all users.

Useful 0 Not Useful 0

Impact Analysis

The vulnerability can have severe impacts as it enables attackers to exfiltrate sensitive data from all users.

• Attackers can steal prompts submitted by users.
• Uploaded documents can be accessed and exfiltrated.
• Extracted entities and knowledge graph content from all users can be compromised.

Overall, this leads to a high risk of data leakage and loss of confidentiality across the entire application instance.

Useful 0 Not Useful 0

Compliance Impact

This vulnerability allows unauthenticated attackers to overwrite global configuration and redirect all LLM operations to attacker-controlled endpoints, enabling exfiltration of sensitive data such as prompts, uploaded documents, extracted entities, and knowledge graph content from all users.

Such unauthorized data access and exfiltration could lead to violations of data protection regulations like GDPR and HIPAA, which require strict controls over personal and sensitive information to prevent unauthorized disclosure.

Therefore, exploitation of this vulnerability may result in non-compliance with these standards due to the compromise of confidentiality and integrity of protected data.

Useful 0 Not Useful 0

Detection Guidance

This vulnerability can be detected by monitoring for unauthorized account registrations and suspicious modifications to the global LLM provider configuration via the /api/v1/settings endpoint. Since the vulnerability allows unauthenticated attackers to self-register and change settings without admin checks, detection involves checking for unexpected new user accounts and unauthorized API calls to the settings endpoint.

Suggested commands or methods include:

• Review logs for new user registrations without email verification or CAPTCHA.
• Monitor API access logs for POST or PUT requests to /api/v1/settings from newly created or suspicious accounts.
• Use network monitoring tools to detect outbound connections to unknown or attacker-controlled LLM endpoints.
• Example command to check recent user registrations (assuming Linux and typical log location): `grep 'user registration' /var/log/cognee/access.log | tail -n 20`
• Example command to check settings endpoint modifications: `grep '/api/v1/settings' /var/log/cognee/access.log | grep 'POST\|PUT'`

Useful 0 Not Useful 0

Mitigation Strategies

Immediate mitigation steps include:

• Disable public user registration to prevent unauthenticated attackers from creating accounts.
• Restrict access to the /api/v1/settings endpoint to only superusers or administrators by enforcing proper authorization checks.
• Require email verification, CAPTCHA, or admin approval for new user registrations to prevent automated or malicious account creation.
• Apply the security patches or upgrade to Cognee version 1.2.0 or later, where these issues have been addressed.
• Mask API keys in any settings responses to avoid leaking sensitive credentials.
• Implement rate limiting on registration endpoints and audit logging for changes to global settings.

Useful 0 Not Useful 0