Milvus Authentication Bypass Vulnerability on TCP Port 9091

Vulnerability

A critical authentication bypass vulnerability has been identified in Milvus, an open-source vector database, prior to versions 2.5.27 and 2.6.10. The vulnerability arises because Milvus exposes TCP port 9091 by default, allowing unauthenticated access to the full REST API registered on this port. This includes access to all business operations, such as data manipulation and credential management. Additionally, the '/expr' debug endpoint on port 9091 uses a weak, predictable default authentication token, enabling arbitrary expression evaluation that could lead to information disclosure, denial-of-service, or even remote code execution.

Impact

Exploitation of this vulnerability allows for unauthenticated access to the REST API on port 9091, bypassing all application-level access controls. This could lead to unauthorized data manipulation, credential management, and exploitation of the '/expr' endpoint for arbitrary expression evaluation, with potential consequences including information disclosure, denial-of-service, and remote code execution.

Reproduction

The vulnerability can be reproduced by accessing the '/expr' endpoint on port 9091 with the default authentication token. This allows for the evaluation of arbitrary internal Go expressions. Alternatively, the vulnerability can be demonstrated by accessing the REST API on port 9091 without any authentication, enabling unauthorized operations such as data manipulation and credential management.

Remediation

Users are advised to upgrade to Milvus versions 2.5.27 or 2.6.10, where this vulnerability has been patched. In addition, it is recommended to block external access to port 9091 using firewall rules or network policies, and to change the 'etcd.rootPath' from the default value 'by-dev' to a strong, random value.

Added: Feb 13, 2026, 8:01 PM
Updated: Feb 13, 2026, 10:08 PM

Vulnerability Rating

Custom Algorithm
spread
0.0
impact
5.6
exploitability
8.3
remediation
0.0
relevance
3.0
threat
6.4
urgency
2.9
incentive
4.2

Our algorithm analyzes dozens of metrics to generate these 8 key vulnerability categories, which are then combined to calculate the overall risk score.