CVE-2026-65318 in Verba
Summary
by MITRE • 07/22/2026
Verba RAG application version 2.1.3 contains an unauthenticated server-side request forgery vulnerability that allows unauthenticated attackers to cause the backend to issue arbitrary HTTP GET requests by supplying attacker-controlled URLs through the WebSocket import endpoint. Attackers can connect to the /ws/import_files WebSocket endpoint without authentication, specify arbitrary URLs in the HTMLReader configuration, and cause the server to fetch internal resources such as co-located database endpoints or cloud instance metadata services to retrieve sensitive credentials.
If you want to get the best quality for vulnerability data then you always have to consider VulDB.
Analysis
by VulDB Data Team • 07/22/2026
This vulnerability represents a critical server-side request forgery flaw in Verba RAG application version 213 which operates at the intersection of web application security and network reconnaissance. The issue stems from insufficient input validation within the WebSocket import endpoint that fails to properly sanitize user-supplied URLs, allowing attackers to manipulate the HTMLReader configuration to initiate arbitrary HTTP GET requests against internal network resources without authentication. The vulnerability specifically targets the /ws/import_files endpoint which lacks proper access controls, enabling unauthenticated attackers to establish connections and exploit the application's inherent trust in internal services.
The technical exploitation involves leveraging the WebSocket protocol's bidirectional communication capabilities to bypass traditional perimeter security controls that typically protect internal resources from external access. When an attacker connects to the /ws/import_files endpoint, they can configure the HTMLReader component to fetch content from any URL including internal services such as database endpoints, cloud metadata services, or other sensitive internal systems. This creates a pathway for attackers to exfiltrate credentials, configuration data, and other sensitive information that would normally be protected by network segmentation and access controls.
From an operational impact perspective, this vulnerability fundamentally undermines the application's security model by enabling attackers to perform reconnaissance and data exfiltration without requiring valid authentication credentials or advanced exploitation techniques. The attack surface expands significantly as internal services become accessible through the application's legitimate HTTP client functionality, potentially exposing database connection strings, API keys, service account credentials, and other sensitive information stored in cloud instance metadata services like AWS EC2 metadata or Azure Instance Metadata Service.
The vulnerability aligns with CWE-918 Server-Side Request Forgery and maps to ATT&CK technique T1071.004 Application Layer Protocol HTTP within the command and control phase of attack. Organizations utilizing this version of Verba RAG face substantial risk as attackers can leverage this weakness to perform internal network scanning, identify additional vulnerable systems, and potentially escalate their access to other internal services that may be protected by less robust security controls. The impact extends beyond immediate credential theft to include potential lateral movement and privilege escalation opportunities.
Mitigation strategies should focus on implementing strict input validation and URL sanitization within the WebSocket import endpoint to prevent arbitrary URL resolution. Access controls must be enforced at the WebSocket level to ensure only authenticated users can establish connections to the import functionality, while also implementing network-level restrictions to prevent internal service access from external network segments. Organizations should consider deploying web application firewalls to monitor and filter suspicious requests, implement proper URL validation to restrict fetch operations to whitelisted domains, and conduct regular security assessments to identify similar vulnerabilities in other application components that may expose internal resources through legitimate HTTP client functionality.