CVE-2025-23169

Versa Director exposes several services by default, including SSH and PostgreSQL, and uses default credentials for multiple accounts (most with sudo access). This allows attackers to easily gain access. Versa Networks is not aware of exploitation, but a proof of concept has been disclosed.

Versa Director by default uses a common password across all PostgreSQL instances and configures the database to listen on all network interfaces. This allows an unauthenticated attacker to access and administer the database or read local filesystem contents, potentially leading to privilege escalation.

In Versa Director, a command injection vulnerability was detected that allows execution of arbitrary commands on the host operating system via a vulnerable application. The attack is possible when the application passes unsafe user-supplied data to a system shell, largely due to insufficient input validation.

The Versa Director SD-WAN orchestration platform contains an argument injection vulnerability in the Java code handling file uploads. An attacker can bypass MIME type validation and upload arbitrary file types, allowing placement of a malicious file on disk.

The Versa Director SD-WAN orchestration platform includes functionality to initiate SSH sessions to remote CPEs and the Director shell via Shell-In-A-Box. The underlying Python script, shell-connect.py, is vulnerable to command injection through the user argument. This allows an attacker to execute arbitrary commands on the system. Versa Networks is not aware of any reported instance where this vulnerability was exploited. Proof of concept has been disclosed by third party security researchers.

Versa Director offers REST APIs for orchestration and management. Some APIs, such as login screen, banner display, and device registration, do not require authentication. It was discovered that for Directors directly connected to the Internet, one of these APIs can be exploited by injecting invalid arguments into a GET request, potentially exposing authentication tokens of other logged-in users. These tokens can then be used to invoke additional APIs on port 9183. This exploit does not disclose usernames or passwords.