Most recent
navigate open esc close Corpus index built 2026-06-07 23:58 UTC

← All engagement records

Case study May 1, 2026

Thailand Ministry of Public Health: Unauthenticated Ollama with Vision Model

Sector
Critical Infrastructure

NuClide Research · 2026-05-01


Summary

Thailand Ministry of Public Health server running Ollama with 5 models including IBM Granite Vision 2B. Raw Ollama port publicly accessible, no authentication. No cloud proxy. Sector: healthcare / government critical infrastructure.


Infrastructure

FieldValue
IP203.157.41.151
OrgMinistry of Public Health, Thailand
SectorHealthcare, National Government
CountryThailand
Open ports11434 (Ollama, public)

Models

ModelSizeNotes
qwen3.6:35b22 GBGeneral
llama3.2:3b1 GBGeneral
smollm2:135m0 GBLightweight
granite3.2-vision:2b2 GBIBM vision model
gemma3:4b3 GBGeneral

granite3.2-vision:2b system prompt: generic assistant configuration (not domain-specific). If this model processes medical images or documents in any ministerial workflow, model injection could affect those outputs.


Findings

F1: Unauthenticated Ollama on Government Healthcare Infrastructure (CRITICAL)

Port 11434 on Ministry of Public Health infrastructure is publicly accessible. All models injectable via CVE-2025-63389.

F2: Vision Model Injection Surface (HIGH)

granite3.2-vision:2b can process images. If connected to any document or imaging workflow within the ministry, injected system prompts redirect the model’s behavior on visual inputs.


Remediation

OLLAMA_HOST=127.0.0.1:11434
systemctl restart ollama

Disclosure

  • Discovered: 2026-05-01
  • Status: Pending, outreach to Thai NCSB (National Cyber Security Agency of Thailand) / MOPH IT