175,000 Ollama LLM servers exposed online without authentication

A significant number of Ollama servers, estimated at around 175,000 globally, are exposed to the internet without any authentication, posing a security risk. Researchers from Cisco Talos and SentinelOne initially found thousands of unsecured servers, with SentinelOne later identifying a much larger scale of exposed instances, many of which also feature tool-calling capabilities. While Ollama's default configuration is secure for local use, expanding access without proper security measures, such as port forwarding or UPnP, can inadvertently expose these LLM infrastructures. The project's issue tracker shows a long-standing request for built-in authentication, but it remains unaddressed, emphasizing the need for users to implement their own access controls. AI

IMPACT Highlights critical security risks for users deploying LLMs locally, emphasizing the need for robust access controls and authentication.

RANK_REASON The article discusses the security implications and setup of a specific tool (Ollama WebUI) for running LLMs locally, rather than a new frontier model release or significant industry-wide event.

Read on dev.to — LLM tag →

• Censys
• ChatGPT
• Cisco Talos
• GitHub
• Ollama
• Open-WebUI
• SentinelLABS
• SentinelOne
• Shodan
• Universal Plug and Play

• infra
• product
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →