Skip to main content
Mallory
CriticalExploited in the wildCISA KEVPublic exploitVulnerability

CVE-2025-62593: Browser-guard bypass leading to remote code execution in Ray

Published Nov 26, 2025 · Updated 1 month ago
Trending this weekNot trendingOutside this week’s top 100
Mention activity · 30 daysNo mention activity in the last 30 days.
CVSS9.4Critical · v4
EPSS60.2%99th percentile
CISA KEVYesAdded Aug 17, 2026
In the wildActivePer CISA KEV
Vendor
Anyscale
IdentifiersCVE-2025-62593CWE-94 · Improper Control of Generation of…

CVE-2025-62593 is a browser-mediated remote code execution vulnerability affecting Ray versions earlier than 2.52.0. Ray Dashboard's browser-request middleware identifies browser traffic by checking whether the User-Agent header begins with "Mozilla." Firefox and Safari allow attacker-controlled JavaScript to modify this header, bypassing the restriction. Combined with DNS rebinding, a malicious webpage or advertisement can make a developer's browser submit malicious jobs to a local or network-accessible, unauthenticated Ray service. Submitted jobs execute arbitrary commands with the privileges of the Ray process. Ray 2.52.0 fixes the vulnerability.

Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Analyst brief

Impact, mitigation & remediation

What it means and what to do now: patch path, interim mitigations, and the assume-compromise checklist.

Impact

What an attacker gets, and what they’ve been doing with it.

Successful exploitation permits arbitrary shell command execution on affected development or testing systems and reachable internal Ray instances, under the Ray process's privileges. This can expose source code, credentials, training data, and model artifacts, enable unauthorized CPU or GPU workloads, and facilitate lateral movement into accessible internal services or cloud infrastructure. Exploitation in the wild has been confirmed, although reported RondoDox attempts include an ineffective User-Agent choice that failed to bypass the browser guard.

Mitigation

If you can’t patch tonight, do this now.

Restrict Ray Dashboard and API access using authenticated proxies, firewall rules, and network segmentation. Enable Ray's optional token authentication, introduced in 2.52.0 and disabled by default. Use DNS-rebinding protections and block unnecessary Dashboard access. Loopback binding alone does not prevent this browser-mediated attack. If effective mitigations cannot be applied, discontinue affected deployments until they can be securely upgraded.

Remediation

Patch, then assume compromise.

Upgrade Ray to version 2.52.0 or later. Version 2.52.1 adds stronger browser-header testing for defense in depth. Rebuild container and deployment images containing vulnerable Ray dependencies and verify that deployed environments no longer retain older releases. Review affected systems for unexpected job submissions and unapproved commands executed by Ray processes.

Public exploits

Exploits

1 exploit validated by Mallory's analysis. Hidden: 4 rejected by Mallory's analysis as not a working exploit.

View more in app →
CVE-2025-62593-PoCThis repository is a small standalone Python proof-of-concept exploit for CVE-2025-62593 affecting Ray. It contains three files: a .gitignore, a README describing the vulnerability and usage, and the main exploit script ray_rce_poc.py. The code is stdlib-only and intended to directly exercise Ray Dashboard HTTP API endpoints rather than relying on the browser-based DNS rebinding scenario mentioned in the README. The exploit workflow is straightforward and operational: it accepts either a single target URL or a file of targets, builds a malicious Ray job entrypoint, submits that job to POST /api/jobs/, polls GET /api/jobs/ until the submitted job reaches a terminal state, and then fetches GET /api/jobs/<submission_id>/logs to recover command output. The payload is attacker-controlled via --command and is wrapped inside a Python one-liner that executes the shell command on the Ray head node, redirects stdout/stderr to a randomized file under /tmp, and prints the captured contents back into the Ray logs. The script includes basic vulnerability triage logic: HTTP 401/403/404 responses are treated as not vulnerable or inaccessible, while HTTP 200 with a returned submission_id is treated as successful job submission. It is not merely a detector because it actively executes arbitrary commands and confirms RCE by retrieving output. It supports batch scanning from a targets file and can write a JSON summary of results. Overall, the repository’s purpose is to provide a practical unauthenticated remote code execution PoC against exposed Ray Dashboard instances on port 8265.Boreas37 · Disclosed Aug 18, 2026 · python · markdown · network · webOperational
Exposure surface

Affected products & vendors

Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.

VendorProductType
AnyscaleRayapplication
FAQ

Frequently asked questions

Is CVE-2025-62593 exploited in the wild?
Yes. As of Oct 9, 2026, Mallory has reports of exploitation in the wild, and CISA added it to KEV on Aug 17, 2026.
Is there a public exploit for CVE-2025-62593?
Yes. Mallory tracks 1 public exploit or proof of concept for CVE-2025-62593, the earliest disclosed Aug 18, 2026.
Is there a patch for CVE-2025-62593?
Upgrade Ray to version 2.52.0 or later. Version 2.52.1 adds stronger browser-header testing for defense in depth.
What products are affected by CVE-2025-62593?
Anyscale Ray.
How severe is CVE-2025-62593?
CVE-2025-62593 has a CVSS v4 score of 9.4 (Critical) and an EPSS score of 60.2% (99th percentile) as of Oct 9, 2026.
Activity feed

Recent activity

92 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.

View more in app →
nuclei templates pull requestsNews · Sep 12, 2026[CVE-2025-62593] Ray - Remote Code Execution by kiyx · Pull Request #17232 · projectdiscovery/nuclei-templates · GitHubA critical remote-code-execution vulnerability affecting Ray, for which the content describes a Nuclei HTTP detection template targeting an exposed Ray dashboard service on port 8265.Read more ↗nuclei templates pull requestsNews · Sep 8, 2026Add template for CVE-2025-62593 (Ray RCE) by Tito0015 · Pull Request #17171 · projectdiscovery/nuclei-templates · GitHubAn unauthenticated remote code execution vulnerability affecting the Ray API. The content describes a Nuclei detection template targeting Ray job API endpoints, including POST /api/jobs/ and GET /api/jobs/{{jobid}}/logs.Read more ↗nuclei templates pull requestsNews · Sep 8, 2026Add CVE-2025-62593 template: Ray Dashboard browser-guard bypass to RCE by mohamed-bashir-dev · Pull Request #17179 · projectdiscovery/nuclei-templates · GitHubA browser-guard bypass vulnerability in the Ray Dashboard that can lead to remote code execution (RCE). The content references Ray versions 2.51.1 and 2.52.0 in the context of a Nuclei template, but does not establish affected or fixed version boundaries.Read more ↗tenable nessus pluginsNews · Sep 1, 2026Ray < 2.52.0 Remote Code Execution (CVE-2025-62593) | Tenable®A high-severity vulnerability affecting Anyscale Ray, identified as CVE-2025-62593. The provided CVSS vectors indicate network reachability, low attack complexity, no required privileges, user interaction, and high impacts to confidentiality, integrity, and availability.Read more ↗
+15 more in the timeline. Sign up to see them
Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Mallory platform

The version that knows your environment.

This page is what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which adversaries are exploiting it right now, and which detections to deploy tonight.
Exposure mappingQuery your assets running an affected version, and investigate the blast radius.
Threat actor evidenceEvery report linking this CVE to a named adversary.
Associated malware9Malware families linked to this CVE, with evidence and IOCs.
Detection signatures1YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Vendor-by-vendor mappingCross-references every affected SKU, including bundled OEM variants.
Social activity69Community discussion across Reddit, Mastodon, and other social sources.