ShadowRay Active Exploitation of Exposed Ray Clusters
Ray
Threat actors actively exploited unauthenticated and exposed Ray clusters over several months for remote code execution. The attacks compromised significant computing resources and valuable enterprise data.
- CVSS
- 9.8
- EPSS
- 0.84 · top 0.3% · 2026-11-13
- fix status
- Fix status unknown
- first report
- 28 Nov 2023
- last update
- 14 Mar 2025
- CVE ids
Timeline · 6
- 28 Nov 2023CVE-2023-48022: Ray has arbitrary code execution via jobs submission API
- 14 Mar 2025ShadowRay: Hijacking Exposed Ray Clusters
- 14 Mar 2025Anyscale Update on CVEs
- 14 Mar 2025CVE-2023-48022
- 14 Mar 2025ShadowRay: First Known Attack Campaign Targeting AI Workloads Actively Exploited In The Wild
- 14 Mar 2025ShadowRay: AI Infrastructure Is Being Exploited In the Wild
Sources · 6
Advisories and databases
- 2023-11-28CVE-2023-48022: Ray has arbitrary code execution via jobs submission APIGitHub Advisory Database · github.com/advisories/GHSA-6wgj-66m2-xxp2
- 2025-03-14CVE-2023-48022NVD · nvd.nist.gov/vuln/detail/CVE-2023-48022
News and write-ups
- 2025-03-14Anyscale Update on CVEsanyscale.com · anyscale.com/blog/update-on-ray-cves-cve-2023-6019-cve-2023-6020-cve-2023-6021-c
- 2025-03-14ShadowRay: First Known Attack Campaign Targeting AI Workloads Actively Exploited In The Wildoligo.security · oligo.security/blog/shadowray-attack-ai-workloads-actively-exploited-in-the-wild
- 2025-03-14ShadowRay: AI Infrastructure Is Being Exploited In the WildProtect AI · protectai.com/threat-research/shadowray-ai-infrastructure-is-being-exploited-in-
Research and catalogues
- 2025-03-14ShadowRay: Hijacking Exposed Ray ClustersMITRE ATLAS · atlas.mitre.org/studies/AML.CS0023
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Patch now
- AI role
- Attack on AI
- categories
- Infrastructure vulnerabilityAgent hijack
- severity
- 3 · Confirmed exploitation
- fix status
- Fix status unknown
- targets
- Framework / tooling
- owasp llm top 10
- LLM03:2025 Supply ChainLLM06:2025 Excessive Agency
- mitre atlas
- AML.T0053 AI Agent Tool Invocation
- dates
- happened 5 Sep 2023 · first reported 28 Nov 2023
- origin
- MITRE ATLASDerived from MITRE ATLAS (Apache 2.0, © MITRE).
- record
- confidence 0.95 · created 2026-10-18 06:55Z · updated 2026-10-18 07:00Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# ShadowRay Active Exploitation of Exposed Ray Clusters https://gcdradar.vercel.app/incidents/shadowray-active-exploitation-of-exposed-ray-clusters · AIR-2025-0049 · closed 2025-03-28 - Lane: Patch now - Severity: 3 (Confirmed exploitation) - Fix status: Fix status unknown - First reported: 2023-11-28 (happened 2023-09-05) - AI role: Attack on AI - Products: Ray - CVE ids: CVE-2023-48022 - CVSS: 9.8 ## Summary Threat actors actively exploited unauthenticated and exposed Ray clusters over several months for remote code execution. The attacks compromised significant computing resources and valuable enterprise data. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2025-03-14: ShadowRay: AI Infrastructure Is Being Exploited In the Wild (https://protectai.com/threat-research/shadowray-ai-infrastructure-is-being-exploited-in-the-wild) - 2025-03-14: ShadowRay: First Known Attack Campaign Targeting AI Workloads Actively Exploited In The Wild (https://oligo.security/blog/shadowray-attack-ai-workloads-actively-exploited-in-the-wild) - 2025-03-14: CVE-2023-48022 (https://nvd.nist.gov/vuln/detail/CVE-2023-48022) - 2025-03-14: Anyscale Update on CVEs (https://anyscale.com/blog/update-on-ray-cves-cve-2023-6019-cve-2023-6020-cve-2023-6021-cve-2023-48022-cve-2023-48023) - 2025-03-14: ShadowRay: Hijacking Exposed Ray Clusters (https://atlas.mitre.org/studies/AML.CS0023) ## Sources - 2023-11-28 GitHub Advisory Database: CVE-2023-48022: Ray has arbitrary code execution via jobs submission API https://github.com/advisories/GHSA-6wgj-66m2-xxp2 - 2025-03-14 MITRE ATLAS: ShadowRay: Hijacking Exposed Ray Clusters https://atlas.mitre.org/studies/AML.CS0023 - 2025-03-14 anyscale.com: Anyscale Update on CVEs https://anyscale.com/blog/update-on-ray-cves-cve-2023-6019-cve-2023-6020-cve-2023-6021-cve-2023-48022-cve-2023-48023 - 2025-03-14 NVD: CVE-2023-48022 https://nvd.nist.gov/vuln/detail/CVE-2023-48022 - 2025-03-14 oligo.security: ShadowRay: First Known Attack Campaign Targeting AI Workloads Actively Exploited In The Wild https://oligo.security/blog/shadowray-attack-ai-workloads-actively-exploited-in-the-wild - 2025-03-14 Protect AI: ShadowRay: AI Infrastructure Is Being Exploited In the Wild https://protectai.com/threat-research/shadowray-ai-infrastructure-is-being-exploited-in-the-wild Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closed8.6Gradio CVE-2023-51449: File traversal and SSRF vulnerability in Gradio
Patch now · Disclosed · since 21 Dec 2023 · updated 22 Dec 2023 · CVE-2023-51449
AI - closed7.8Hugging Face libraries CVE-2023-7018: Deserialization vulnerability in Hugging Face transformers
Patch now · Disclosed · since 20 Dec 2023 · CVE-2023-7018
AI - closed9.0Hugging Face libraries CVE-2023-6730: Deserialization vulnerability in Hugging Face transformers
Patch now · Disclosed · since 19 Dec 2023 · CVE-2023-6730
AI - closed9.6Gradio CVE-2023-6572: Information exposure vulnerability in Gradio
Patch now · Disclosed · since 14 Dec 2023 · CVE-2023-6572
AI - closed7.2LiteLLM CVE-2024-4264: LiteLLM Unsafe Eval Remote Code Execution
Patch now · Disclosed · since 18 May 2024 · CVE-2024-4264
AI - closed9.0LoLLMs CVE-2024-2366: lollms-webui Remote Code Execution via Path Traversal
Patch now · Disclosed · since 16 May 2024 · CVE-2024-2366
AI