https://github.com/5GSEC/MobiLLM
Domain-specific LLM for cellular security
https://github.com/5GSEC/MobiLLM
Last synced: about 1 month ago
JSON representation
Domain-specific LLM for cellular security
- Host: GitHub
- URL: https://github.com/5GSEC/MobiLLM
- Owner: 5GSEC
- License: apache-2.0
- Created: 2025-08-19T14:18:39.000Z (10 months ago)
- Default Branch: main
- Last Pushed: 2025-09-16T15:20:29.000Z (9 months ago)
- Last Synced: 2025-10-12T16:45:15.397Z (8 months ago)
- Language: Python
- Size: 689 KB
- Stars: 1
- Watchers: 0
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-telco - MobiLLM - 10]` - Domain-specific LLM for cellular security from the 5GSEC group. Trained for analyzing and reasoning about 5G/cellular security incidents and threats. (AI & Machine Learning / Telecom LLMs & AI Assistants)
README
# MobiLLM
**Domain-specific LLM for cellular security**
MobiLLM is an intelligent multi-agent system designed specifically for 5G network security analysis and response. It leverages Large Language Models (LLMs) to provide automated threat detection, analysis, and mitigation strategies for cellular networks.
## 🚀 Features
- **Multi-Agent Architecture**: Specialized agents for different security tasks
- **5G Network Security**: Domain-specific knowledge for cellular security
- **MITRE ATT&CK Integration**: Maps threats to standard attack frameworks
- **Automated Response**: Suggests and executes countermeasures
- **Real-time Analysis**: Processes network events and telemetry data
- **Human-in-the-Loop**: Approval system for critical actions
## 🏗️ Architecture
MobiLLM uses a multi-agent architecture built with LangGraph:
### Agent Specializations
- **Chat Agent**: General network queries and service management
- **Security Analysis Agent**: Threat detection and analysis
- **Classification Agent**: MITRE ATT&CK technique mapping
- **Response Planning Agent**: Countermeasure strategy development
- **Config Tuning Agent**: Network configuration modifications
## 📋 Prerequisites
- Python 3.8+
- Google Gemini API key
- Docker (for xApp deployment)
- Kubernetes cluster (for production deployment)
## 🛠️ Installation
### 1. Clone the Repository
```bash
git clone https://github.com/5GSEC/MobiLLM.git
cd MobiLLM
```
### 2. Install Dependencies
```bash
pip install -r requirements.txt
```
### 3. Set Environment Variables
```bash
export GOOGLE_API_KEY="your_gemini_api_key_here"
```
### 4. Verify Installation
```bash
python -c "from MobiLLM import MobiLLM_Multiagent; print('Installation successful!')"
```
## 🔧 Configuration
### Environment Variables
| Variable | Description | Default |
|----------|-------------|---------|
| `GOOGLE_API_KEY` | Google Gemini API key | Required |
| `XAPP_ROOT_PATH` | xApp root directory | `./xApp` |
| `OAI_RAN_CU_CONFIG_PATH` | OAI RAN CU config path | Optional |
### Sample Data
The system includes sample 5G network data for testing:
- UE MobiFlow data
- Base Station MobiFlow data
- Security event data (MobieXpert, MobiWatch)
- Service status data
## 🚀 Quick Start
### Basic Usage
```python
from MobiLLM import MobiLLM_Multiagent
# Initialize the agent
agent = MobiLLM_Multiagent()
# Chat with the agent
response = agent.chat("How many UEs are connected to the network?")
print(response["output"])
# Security analysis
result = agent.security_analysis("Analyze event ID 123 for security threats")
print(result["output"])
```
### Security Analysis Example
```python
# Conduct security analysis
result = agent.invoke("""[security analysis]
Event Details:
- Source: MobieXpert
- Name: RRC Null Cipher
- Cell ID: 20000
- UE ID: 54649
- Severity: Critical
- Description: UE uses null cipher mode in RRC session
""")
print("Threat Summary:", result["threat_summary"])
print("MITRE Techniques:", result["mitre_technique"])
print("Countermeasures:", result["countermeasures"])
```
## 🧪 Testing
### Run Test Suite
```bash
# Run all tests
python -m MobiLLM.test.test_llm
# Run baseline tests
python -m MobiLLM.test.baseline
```
### Test Individual Components
```bash
# Test LLM functionality
python -m MobiLLM.test_llm
# Test specific agents
python -c "
from MobiLLM import MobiLLM_Multiagent
agent = MobiLLM_Multiagent()
print('Agent initialized successfully')
"
```