NuClide Research
Most recent
↑↓ navigate ↵ open esc close Corpus index built 2026-06-07 23:58 UTC

← Research library

CRITICAL · Disclosure May 1, 2026

Tw Ncku

Bill Rondell, Nick Adams & Dade Murphy — May 1, 2026 — critical · disclosure — 3 min read — 552 words

To: mailservice@ncku.edu.tw Subject: Unauthenticated AI inference endpoint, National Cheng Kung University (140.116.158.98)

Nicholas Michael Kloster / NuClide Research nicholas@nuclide-research.com

2026-05-01

Re: Unauthenticated Ollama AI inference endpoint, National Cheng Kung University IP / Host: 140.116.158.98 Severity: CRITICAL

I’m an independent security researcher. I hold CISA disclosures CVE-2025-4364 and ICSA-25-140-11 and conduct good-faith AI infrastructure research under the NuClide Research umbrella. This is an unsolicited disclosure, no engagement exists with your organization, and I have not accessed, modified, or exfiltrated any data beyond what was necessary to confirm the exposure.

Summary

National Cheng Kung University (NCKU), one of Taiwan’s top engineering universities, has an Ollama instance running on non-standard port 22222. The MiniMax cloud proxy leaks the Ollama Connect account nckusoc-3090, indicating NCKU School of Computing (SOC) department server with an NVIDIA RTX 3090 GPU.

Infrastructure

FieldValue
IP140.116.158.98
rDNS,
OrgMinistry of Education Computer Center (TANET)
InstitutionNational Cheng Kung University, SOC Department
CountryTaiwan
Open ports22222 (Ollama non-standard port, public)

Note: IP routes through Taiwan’s Ministry of Education TANET network, shared by major Taiwanese universities.

Models

ModelSizeTypeCred Leak
minimax-m2.7:cloud0 GB☁️ Cloud proxynckusoc-3090
qwen3.6:35b22 GBLocal,
gpt-oss:20b12 GBLocal,
mistral-small3.2:24b14 GBLocal,
gemma3:27b16 GBLocal,
gemma3:12b7 GBLocal,
gemma3:4b3 GBLocal,
llama3.2:3b1 GBLocal,

Credential Leak

{
  "error": "unauthorized",
  "signin_url": "https://ollama.com/connect?name=nckusoc-3090&key=<base64>"
}
  • Username: nckusoc-3090
  • SSH Public Key: ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIMwH+iskAm2POkZim1R1+IHud67QvLGpB7DRs19xh/pb

nckusoc = NCKU School of Computing; 3090 = NVIDIA RTX 3090 GPU identifier.

Findings

F1, Credential Leak via Non-Standard Port (HIGH): Ollama running on port 22222 instead of default 11434. MiniMax cloud proxy leaks nckusoc-3090 account credentials.

F2, Model Injection (HIGH): All 8 models injectable via CVE-2025-63389.

Taiwan MOE TANET Context

Multiple Taiwanese universities share the MOE TANET (Taiwan Academic Network) IP space (140.112.x.x - NTU, 140.114.x.x - NTHU, 140.116.x.x - NCKU, 140.136.x.x - FJU). Ollama instances observed across this network on 2026-05-01:

IPInstitutionModelsCloud
140.112.91.82NTU (Electrical Engineering)4minimax-m2.7
140.112.18.214NTU (PC-214)5,
140.112.233.108NTU (GPU cluster g1)11,
140.116.82.105NCKU / TANET8deepseek-v4-pro
140.116.158.98NCKU (SOC-3090)8minimax-m2.7
140.136.192.220FJU (Medical Public Health)8,
140.136.239.75FJU (net2net)5,
163.25.105.115TANET node9,
163.13.202.114TANET node2,
140.136.149.212TANET node2,
210.70.138.233TANET node3,

Why it matters

The credential leak (username + SSH public key) exposes your service account to enumeration and credential-stuffing against other services. Medical AI models exposed without authentication create compliance risk (potential HIPAA/patient-data adjacent exposure depending on RAG content).

One-line fix

OLLAMA_HOST=127.0.0.1:11434
systemctl restart ollama

This rebinds Ollama to loopback only. If running in Docker: docker run -p 127.0.0.1:11434:11434 ollama/ollama.

CVE-2025-63389

All models on this instance are injectable via the unauthenticated /api/create endpoint, an attacker can overwrite any model’s system prompt or delete models entirely. No patch exists as of this disclosure.

Reference

Full technical details, parameter counts, and remediation notes are in this public research repository: AI-LLM-Infrastructure-OSINT/blob/main/case-studies/universities/TW/ncku.md

This research is part of a broader sweep of university AI infrastructure exposures documented at: AI-LLM-Infrastructure-OSINT/blob/main/case-studies/universities/OVERVIEW.md

I’m happy to answer questions or assist with verification. No response is required.

Regards, Nicholas Michael Kloster / NuClide Research nicholas@nuclide-research.com AI-LLM-Infrastructure-OSINT