/
← Voltar ao feed
high

Trend Micro reports thousands of unauthenticated AI servers exposed online, plus flaws in ChromaDB, Redis, NVIDIA Triton and Container Toolkit

1 fonte(s) · 1 atualização(ões) · publicado 29/07/2025 09:00 · coletado há 1 dia

Trend Micro published its State of AI Security Report, 1H 2025, warning that AI infrastructure is being built from unsecured and unpatched components. The research found 200+ ChromaDB servers, 2,000 Redis servers and 10,000+ Ollama servers exposed to the internet with no authentication. It also reports zero-day vulnerabilities and exploits in ChromaDB, Redis, NVIDIA Triton and NVIDIA Container Toolkit. Trend warns of risks including data theft, poisoning and ransom. The article does not describe confirmed in-the-wild exploitation or name any threat actor.

Technical Details

  • Core AI components (ChromaDB, Redis, NVIDIA Triton, NVIDIA Container Toolkit) contain zero-day vulnerabilities and exploits.
  • At Pwn2Own Berlin, which featured a new AI category, researchers demonstrated an exploit for the Redis vector database that stemmed from an outdated Lua component.
  • Pwn2Own researchers also demonstrated an exploit for the NVIDIA Container Toolkit. Containerized AI infrastructure inherits cloud and container threats.
  • Rushed deployments lead to accidental internet exposure of AI servers, which adversaries can probe.
  • Open-source libraries in AI frameworks can carry vulnerabilities into production unnoticed.

Remediation

  • Improve patch management and run vulnerability scans.
  • Maintain an inventory of all software components, including third-party libraries and subsystems.
  • Follow container security best practices: minimal base images, input sanitization and runtime monitoring or security tools.
  • Check configurations to ensure AI infrastructure servers are not exposed to the internet and require authentication.

Linha do tempo

  1. newsroom.trendmicro.com ↗ · pub. 29/07/2025 09:00 · coletado 19/10 18:02 primeiro registro

    Trend Micro reports thousands of unauthenticated AI servers exposed online, plus flaws in ChromaDB, Redis, NVIDIA Triton and Container Toolkit

    ler a análise desta fonte

    Trend Micro published its State of AI Security Report, 1H 2025, warning that AI infrastructure is being built from unsecured and unpatched components. The research found 200+ ChromaDB servers, 2,000 Redis servers and 10,000+ Ollama servers exposed to the internet with no authentication. It also reports zero-day vulnerabilities and exploits in ChromaDB, Redis, NVIDIA Triton and NVIDIA Container Toolkit. Trend warns of risks including data theft, poisoning and ransom. The article does not describe confirmed in-the-wild exploitation or name any threat actor.

    Technical Details

    • Core AI components (ChromaDB, Redis, NVIDIA Triton, NVIDIA Container Toolkit) contain zero-day vulnerabilities and exploits.
    • At Pwn2Own Berlin, which featured a new AI category, researchers demonstrated an exploit for the Redis vector database that stemmed from an outdated Lua component.
    • Pwn2Own researchers also demonstrated an exploit for the NVIDIA Container Toolkit. Containerized AI infrastructure inherits cloud and container threats.
    • Rushed deployments lead to accidental internet exposure of AI servers, which adversaries can probe.
    • Open-source libraries in AI frameworks can carry vulnerabilities into production unnoticed.

    Remediation

    • Improve patch management and run vulnerability scans.
    • Maintain an inventory of all software components, including third-party libraries and subsystems.
    • Follow container security best practices: minimal base images, input sanitization and runtime monitoring or security tools.
    • Check configurations to ensure AI infrastructure servers are not exposed to the internet and require authentication.