CISA Warns of Ray-Project Ray Code Injection Vulnerability Exploited in Attacks

Key Takeaways

• The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has issued a warning about active exploitation of a critical code injection vulnerability, CVE-2025-62593, in the Ray-Project Ray framework.
• This flaw enables remote code execution and primarily impacts Ray development environments, particularly when developers are using Firefox or Safari browsers.
• The vulnerability affects Ray versions prior to 2.52.0 and has been addressed in version 2.52.0 and later.
• CISA has added the flaw to the KEV catalog, mandating federal agencies to remediate by August 20, 2026.

Critical Ray-Project Ray Vulnerability Under Active Exploitation

CISA has issued an urgent alert regarding a severe vulnerability, tracked as CVE-2025-62593, within the Ray-Project Ray distributed computing framework. This critical flaw, which permits remote code execution, has been added to CISA’s Known Exploited Vulnerabilities catalog, confirming its active exploitation in the wild.

Understanding the Ray Framework and the Vulnerability

Ray is an open-source distributed computing framework widely adopted by Python developers and artificial intelligence (AI) teams. It is instrumental in scaling machine learning tasks, handling large-scale data processing, and managing complex application workloads.

The vulnerability, designated CVE-2025-62593, originates from inadequate security measures within Ray’s HTTP API endpoints. Specifically, job-related endpoints are susceptible to specially crafted requests that can initiate arbitrary code execution. Ray versions predating 2.52.0 are affected, with the issue being particularly dangerous for developers operating vulnerable Ray instances on their local workstations or development servers while browsing the web with Firefox or Safari.

The Ray Project has since released version 2.52.0, which incorporates a fix for this vulnerability.

Exploitation Mechanism and Attack Scenario

Initially, Ray implemented a defense mechanism to prevent browser-based access by verifying if the User-Agent header began with “Mozilla.” However, this protection proved insufficient and could be bypassed. In Firefox and Safari, attackers can manipulate the User-Agent value through specific browser request behaviors, thereby circumventing the intended security filter.

Attackers can combine this User-Agent manipulation with DNS rebinding techniques. DNS rebinding allows an attacker to trick a victim’s browser into communicating with local or internal services, bypassing typical network security boundaries. A probable attack scenario involves a developer running an unpatched Ray instance who visits a malicious website or encounters a compromised advertisement. The attacker-controlled page could then send requests to the Ray service running on the developer’s machine, leading to arbitrary code execution with the same privileges as the Ray process. The vulnerability is linked to CWE-94 (Improper Control of Code Generation) and CWE-352 (Cross-Site Request Forgery).

Although user interaction is necessary for the attack, the attacker does not require prior access to the target Ray environment. This factor significantly elevates the risk for developers who maintain local Ray deployments while engaging in general web browsing.

CISA’s Mandate and Remediation Timeline

CISA officially added this flaw to its Known Exploited Vulnerabilities (KEV) catalog on August 17, 2026, setting a remediation deadline of August 20, 2026, for federal agencies. While CISA’s listing confirms active exploitation, it has not disclosed details concerning specific threat actors, victims, or the nature of the attacks, such as whether ransomware is involved.

What You Should Do

• Immediate Upgrade: Organizations must prioritize upgrading all Ray installations to version 2.52.0 or newer. This includes developer endpoints, build servers, container images, and cloud-based compute environments.
• Rebuild Images: Rebuild any container images that may contain older, vulnerable Ray dependencies, ensuring that package management tools have not retained an insecure Ray release.
• Network Segmentation and Access Control: If immediate patching is not feasible, ensure that Ray dashboards and APIs are not exposed to untrusted networks. Implement strict access controls through network segmentation, authenticated proxies, VPNs, and robust firewall rules.
• Log Review: Regularly review logs for unusual activity, including unexpected job submissions, suspicious API requests, abnormal browser activity, or unauthorized commands executed by Ray processes.
• Discontinue Use: CISA advises that organizations unable to implement effective mitigations should consider discontinuing the use of affected Ray deployments until they can be securely upgraded.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.