CVE-2025-62593: Ray browser-based RCE via insufficient User-Agent guard

Automated analysis

Generated by VULONE's analysis model from the NVD record, CISA KEV and EPSS data on 17 September 2026. Confidence: high.

critical priorityCVSS 4.0 score is 9.4 critical, the flaw is in CISA KEV with active exploitation, and it enables unauthenticated network-reachable RCE with high impact.

What it is

Ray, an AI compute engine, contains a critical remote code execution flaw before version 2.52.0. Its defense against browser-based attacks relies on checking whether the User-Agent header starts with "Mozilla", which the fetch specification allows attackers to set arbitrarily. Combined with DNS rebinding, a developer running Ray who visits a malicious site or malvertising can be compromised.

Impact

An attacker gains remote code execution on a developer's Ray instance, with high confidentiality, integrity and availability impact across the vulnerable system and subsequent scope. This can lead to full compromise of the developer host and any connected Ray cluster resources.

Attack surface

Reached over the network with no privileges required, but user interaction is required: the victim must visit a malicious website or be served malicious advertising while running Ray. No authentication is needed per the CVSS vector (PR:N).

Exploitation

CISA added this to the Known Exploited Vulnerabilities catalog on 2026-08-17, indicating active exploitation. EPSS gives a 30-day probability of 0.16888 (96.9th percentile), and the vendor advisory is tagged Exploit.

What to do

• Upgrade Ray to version 2.52.0 or later, which contains the patch (commit 70e7c72780bdec075dba6cad1afe0832772bfe09).
• If immediate patching is not possible, follow CISA BOD 26-04 guidance: restrict or discontinue use of exposed Ray services and evaluate each asset's internet exposure.
• Do not rely on User-Agent filtering as a defense; block browser-originated requests to Ray endpoints at the network or proxy layer.
• Bind Ray dashboards and services to localhost or trusted interfaces only, and avoid running Ray on developer workstations while browsing untrusted sites.
• Apply DNS rebinding protections such as DNS pinning or validating Host headers on Ray endpoints.

Detection

• Monitor Ray service logs for requests with unusual or spoofed User-Agent headers, especially those claiming Mozilla origins from unexpected source IPs.
• Alert on DNS queries resolving to internal Ray hostnames or IPs from browser processes, which may indicate DNS rebinding.
• Hunt for unexpected child processes or command execution spawned by Ray worker or dashboard processes.
• Review network flows to Ray ports from external or untrusted networks and flag any exposure of Ray services beyond localhost.

This assessment is produced automatically and is not human-reviewed. Verify against the vendor advisory before acting on it.