CVE-2026-33340 Scanner

LoLLMs WEBUI is a software product used for managing and interfacing with machine learning models. It is utilized by developers and technical users to test and execute various AI and machine learning tasks, often in research or production environments. Typically, it is deployed in environments where manipulating models and accessing remote resources are necessary. The tool provides an interactive user interface for users to manage jobs, tasks, and data inputs efficiently. Due to its use in sensitive operations and data exchanges, secure deployment and management practices are crucial for its operation. The product's popularity stems from its ability to handle complex computations and interact with a variety of data sources seamlessly.

Server-Side-Request-Forgery (SSRF) is a vulnerability that allows an attacker to force a server to connect to unauthorized locations. This type of flaw permits an external party to access the server or initiate connections from the server to internal services or external resources that they should not be able to reach. The flaw might lead the server to disclose sensitive information from internal systems, or even let the attacker perform unauthorized actions. SSRF vulnerabilities can be detrimental by exposing sensitive metadata or allowing the manipulation of network services. Understanding and mitigating this vulnerability is critical to protecting internal networks and data from unwanted exposure and potential abuse.

The vulnerability is found at the endpoint "/api/proxy" of the LoLLMs WEBUI, which does not restrict attempts to access internal servers through crafted requests. An attacker could make unauthenticated access using HTTP POST requests with crafted JSON payloads that direct the server to connect to specified URLs. The control of the request path through this endpoint serves as a mechanism through which these exploits can be tested. Unchecked, this can enable attackers to initiate GET requests to internal IP addresses or other unauthorized targets. The exploitation requires no special privileges or authentication, making it particularly severe if unmitigated.

If exploited, this vulnerability could allow attackers to scan the internal network infrastructure, exfiltrate sensitive data such as cloud-specific metadata, or misuse internal services to create further vulnerabilities. It can lead to unauthorized access to confidential company resources or systems. Also, this flaw might become a stepping stone for broader attacks aimed at the organization's infrastructure and beyond. Ensuring such access pathways are secured effectively is vital to avoid data exposure or unauthorized data manipulations that can impact the organization's security posture.

REFERENCES

• https://github.com/ParisNeo/lollms-webui/security/advisories/GHSA-mcwr-5469-pxj4
• https://nvd.nist.gov/vuln/detail/CVE-2026-33340