CVE-2025-22892 is a high-severity vulnerability affecting Intel's OpenVINO™ model server software versions prior to 2024.4. The vulnerability arises from uncontrolled resource consumption, which can be triggered by an unauthenticated attacker with adjacent network access. Specifically, the flaw allows an attacker to send crafted requests or traffic to the OpenVINO model server, causing it to consume excessive system resources such as CPU, memory, or network bandwidth. This resource exhaustion can lead to a denial of service (DoS) condition, rendering the model server unavailable to legitimate users. The vulnerability does not require user interaction, privileges, or authentication, and it does not impact confidentiality, integrity, or availability beyond the DoS effect. The CVSS 4.0 vector (AV:A/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N) indicates that the attack requires adjacent network access (such as the same local network segment), has low attack complexity, no privileges or user interaction, and results in high impact on availability. No known exploits are currently reported in the wild, and no patches or mitigation links were provided at the time of publication. The vulnerability is relevant to deployments of OpenVINO model server software, which is used for serving AI and machine learning models in production environments, often in edge computing or local network scenarios.

For European organizations, the impact of this vulnerability can be significant, particularly for those relying on OpenVINO model server software for AI inference workloads in critical applications such as manufacturing automation, healthcare diagnostics, smart city infrastructure, and telecommunications. A successful DoS attack could disrupt AI-driven services, causing operational downtime, degraded service quality, and potential financial losses. Since the attack requires only adjacent network access and no authentication, internal threat actors or compromised devices within the same network segment could exploit this vulnerability. This risk is heightened in environments with less network segmentation or where OpenVINO servers are exposed to broader internal networks. Additionally, disruption of AI services could indirectly affect decision-making processes and automated controls, impacting business continuity and safety-critical operations. Although the vulnerability does not expose data or allow code execution, the denial of service impact on availability is critical for organizations with real-time or near-real-time AI inference requirements.

European organizations should prioritize upgrading OpenVINO model server software to version 2024.4 or later as soon as the patch becomes available from Intel. Until patches are applied, organizations should implement strict network segmentation to isolate OpenVINO servers from untrusted or less secure network segments, limiting adjacent access to only authorized devices and users. Deploying network-level controls such as firewalls, VLANs, or zero-trust network access (ZTNA) can reduce the attack surface. Monitoring network traffic to and from OpenVINO servers for unusual patterns or spikes in resource usage can help detect attempted exploitation. Rate limiting or connection throttling at the network or application layer may mitigate resource exhaustion attempts. Additionally, organizations should review and harden configurations of OpenVINO servers to minimize exposure, disable unnecessary services, and ensure logging is enabled for forensic analysis. Incident response plans should include procedures for rapid detection and recovery from DoS conditions affecting AI inference services.