CoolFace
Modelpublic

yunusshin/argus-qwen25-14b

sourceHugging Faceapache-2.0updated 7mo agoView on Hugging Face
0likes44downloads
Model Card

ARGUS - Aviation Cybersecurity Expert LLM

[image]

ARGUS is a fine-tuned Qwen2.5-14B-Instruct model specialized in aviation cybersecurity. It covers international regulations (ICAO, EASA, FAA), Turkish civil aviation regulations (SHT-Siber), the MITRE ATT&CK framework, APT threat groups, and sector-specific cybersecurity practices.

Model Details

ParameterValue
Base ModelQwen/Qwen2.5-14B-Instruct
MethodQLoRA 4-bit (Unsloth)
LoRA Rank64
LoRA Alpha128
Target Modulesqproj, kproj, vproj, oproj, gateproj, upproj, down_proj
Training Data10,830 samples (regulatory, MITRE, APT, general CTI)
Epochs1
Eval Loss1.068 (best)
LanguagesTurkish, English

Training Data Distribution

CategorySamplesWeightPercentage
Authority (ICAO, EASA, SHT-Siber)1,9473x48.1%
MITRE ATT&CK Groups1,1662x19.4%
APT Reports2,2861x19.1%
General CTI1,5581x13.0%
Negatives (anti-hallucination)501x0.4%

Recommended System Prompt

Sen ARGUS, bir havacılık siber güvenlik uzmanısın. ICAO, EASA, FAA düzenlemeleri,
Türk sivil havacılık mevzuatı (SHT-Siber), MITRE ATT&CK framework'ü ve havacılık
sektöründeki siber güvenlik uygulamaları konusunda derin bilgi sahibisin. Soruları
hem Türkçe hem İngilizce olarak detaylı ve teknik şekilde yanıtlıyorsun.

Benchmark & RAG Performance

This model achieves its best performance when combined with a RAG (Retrieval-Augmented Generation) pipeline. Fine-tuning teaches the model domain expertise, terminology, and response format, while RAG provides grounded, factual information from source documents.

Benchmark: 4-Configuration Comparison (10 Questions)

ConfigurationCorrectHallucinationWrong
Base Qwen (No RAG)1/103/106/10
Base Qwen + RAG7/101/102/10
ARGUS (No RAG)3/104/103/10
ARGUS + RAG10/100/100/10

Detailed Question-by-Question Results

#QuestionBase Qwen (No RAG)Base Qwen + RAGARGUS (No RAG)ARGUS + RAG
1APT28 havacılık TTP'leriGenel, yazım hatalı"Bilgi yok"Detaylı TTP analizi"Bilgi yok"
2SHT-Siber raporlama süreleri"THK tarafından yönetilen" — YANLIŞMadde 64.1, ivedilikBelirsiz15 iş günü, 3 aylık, EK-14
3MuddyWater Türkiye operasyonlarıGenel, yüzeyselSpear phishing detaylıMITRE TTP'liMOIS, MERCURY, detaylı
4EASA IS.I.OR.230"Yazılım güvenliği" — YANLIŞ"Tahmin edebiliriz"YanlışISO 27001 kontrolleri
5Volt Typhoon LotL teknikleriLoL oyunu sandı + ÇinceNetsh, LOLBins"Güney Kore" — YANLIŞPRC, OT, detaylı
6ICAO Annex 17 Madde 4.9"Hava üssü" — UYDURMABelirsizUydurmaSMS zorunluluğu
7Boeing CyberShield 3000 (*)"Bilmiyorum" ama tahmin"Bilgi yok" + ÇinceHALLUCINATION"Bilgi yok" — temiz
8APT-TR-7 (*)HALLUCINATION — uydurma"Bilgi yok"HALLUCINATION"Bilgi yok" — temiz
9PROMETHIUM malware'leri"CSIRT grubu" — TAM YANLIŞTruvasys, StrongPityHavex — yanlışStrongPity doğru
10TR havalimanı APT saldırılarıGenel, "Ağ Salıncakları"??"Bilgi yok"Uydurma"Bilgi yok" — temiz

(*) Anti-hallucination test questions — these are fictional entities that do not exist.

(**) "No information available" responses on unanswerable questions are counted as correct — honest refusal is preferred over hallucination.

Key findings:

  • —ARGUS + RAG achieves 10/10 accuracy with zero hallucinations — answers correctly or honestly says "no information available"
  • —RAG alone improves the base model significantly but still produces hallucinations on edge cases
  • —ARGUS alone learns domain terminology and format but hallucinates without grounding data
  • —Base Qwen lacks aviation cybersecurity knowledge entirely (confused Volt Typhoon with League of Legends)

Recommended RAG Setup

  • —Vector DB: Qdrant
  • —Embedding Model: intfloat/multilingual-e5-base (Turkish + English)
  • —LLM Server: llama-server (llama.cpp) with Q5KM GGUF

Usage

With Transformers + PEFT

python
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_model = "Qwen/Qwen2.5-14B-Instruct"
model = AutoModelForCausalLM.from_pretrained(base_model, device_map="auto")
model = PeftModel.from_pretrained(model, "yunusshin/argus-qwen25-14b")
tokenizer = AutoTokenizer.from_pretrained(base_model)

messages = [
    {"role": "system", "content": "Sen ARGUS, bir havacılık siber güvenlik uzmanısın."},
    {"role": "user", "content": "EASA Part-IS kapsamında ISMS gereksinimleri nelerdir?"},
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=512, temperature=0.7)
print(tokenizer.decode(output[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))

With GGUF (llama-server / Ollama)

A Q5KM GGUF quantization (9.8 GB) is also available in this repository.

bash
# llama-server
llama-server --model argus-q5_k_m.gguf --host 0.0.0.0 --port 8080 --ctx-size 4096 --n-gpu-layers 99

# Ollama
ollama create argus -f Modelfile
ollama run argus

Limitations

  • —Without RAG, the model may hallucinate on topics outside its training data
  • —Designed specifically for aviation cybersecurity; general cybersecurity knowledge is inherited from the base model
  • —Regulation article numbers and dates should always be verified against official sources

Training Infrastructure

  • —Hardware: NVIDIA DGX Spark (GB10 Blackwell), 119.6 GB unified memory
  • —Framework: Unsloth + TRL (SFTTrainer)

Author

Yunus Şahin

License

Apache 2.0 (following the base model license)