Red Hat Security Advisory: RHOAI 3.3.3 - Red Hat OpenShift AI

Critical

A Server-Side Request Forgery (SSRF) vulnerability exists in the MediaConnector class within the vLLM project's multimodal feature set. The load_from_url and load_from_url_async methods fetch and process media from user-provided URLs without adequate restrictions on the target hosts. This allows an attacker to coerce the vLLM server into making arbitrary requests to internal network resources.

CWE-918 - Server-Side Request Forgery (SSRF)

A flaw was found in node-forge. This vulnerability allows unauthenticated attackers to bypass downstream cryptographic verifications and security decisions via crafting ASN.1 (Abstract Syntax Notation One) structures to desynchronize schema validations, yielding a semantic divergence.

CWE-179 - Incorrect Behavior Order: Early Validation

A flaw was found in Lodash. A prototype pollution vulnerability in the _.unset and _.omit functions allows an attacker able to control property paths to delete methods from global prototypes. By removing essential functionalities, this can result in a denial of service.

CWE-1321 - Improperly Controlled Modification of Object Prototype Attributes ('Prototype Pollution')

A flaw was found in qs, a module used for parsing query strings. A remote attacker can exploit an improper input validation vulnerability by sending specially crafted HTTP requests that use bracket notation (e.g., `a[]=value`). This bypasses the `arrayLimit` option, which is designed to limit the size of parsed arrays and prevent resource exhaustion. Successful exploitation can lead to memory exhaustion, causing a Denial of Service (DoS) where the application crashes or becomes unresponsive, making the service unavailable to users.

CWE-770 - Allocation of Resources Without Limits or Throttling