Ghost in the Machine

Advanced Interactive Honeypot Deception System

A full-stack cyber-intelligence platform that simulates realistic Linux terminal environments to trap, study, and profile attackers in real-time. Built with two attack surfaces: a web-based fake terminal and a real SSH server powered by AI.

---

Architecture

The system consists of two main attack surfaces:

1. Web Terminal: A Node.js based terminal simulator using Socket.IO.
2. SSH Honeypot: A Paramiko-based SSH server that intercepts commands.

Both surfaces pipe commands through a response engine (AI-based or static fallbacks) and log all activity to MongoDB.

---

Quick Start

Prerequisites

• Node.js >= 18.x
• Python >= 3.10
• MongoDB (optional, has in-memory fallback)
• OpenAI API Key or Ollama (for AI-powered responses)

1. Web Terminal (Node.js)

# Install dependencies
npm install

# Start the server
npm start

# Open in browser:
# Terminal:  http://localhost:3000
# Dashboard: http://localhost:3000/dashboard

2. SSH Honeypot (Python)

# Install Python dependencies
cd ssh_honeypot
pip install -r requirements.txt

# Configure environment
cp .env.example .env
# Edit .env with your API keys

# Start the SSH honeypot
python ssh_server.py

# Connect as an attacker (testing):
ssh admin@localhost -p 2222

3. Run Analysis

cd ssh_honeypot
python run_analysis.py              # Analyze all sessions
python run_analysis.py -s <id>      # Specific session
python run_analysis.py --top 5      # Top 5 riskiest
python run_analysis.py -e report.json  # Export to JSON

---

Modules

• modules/virtualFs.js: Virtual Linux filesystem.
• modules/commandProcessor.js: Command handlers.
• modules/exploitTrap.js: Exploit detection and traps.
• modules/processSimulator.js: Process management simulation.
• modules/networkSimulator.js: Network tool simulation.
• modules/sessionLogger.js: Session tracking and scoring.
• ssh_honeypot/ssh_server.py: Paramiko SSH server entry point.
• ssh_honeypot/ai_engine.py: LLM-based command processing.
• ssh_honeypot/sentiment_analyzer.py: Behavioral analysis engine.

---

Deception Features

• Breadcrumb System: Fake DB credentials, AWS keys, crypto wallets, SSH keys planted throughout
• Exploit Traps: DirtyCow, PwnKit, OverlayFS, Baron Samedit — all "almost work"
• Wall Messages: Paranoid sysadmin broadcasts triggered by attacker activity
• Visual Deception: Fake "TOP SECRET" documents generated on-the-fly for exfiltration attempts
• Credential Harvesting: sudo/su/passwd prompts capture attacker passwords
• Docker/K8s Traps: Fake container escape paths that lead nowhere

Threat Intelligence

• MITRE ATT&CK Mapping: Automatic technique fingerprinting (T1059, T1082, T1003, etc.)
• Sentiment Analysis: VADER with custom cybersecurity lexicon
• Sophistication Scoring: Script kiddie → Beginner → Intermediate → Advanced
• Frustration Detection: Multi-signal analysis (repeated commands, errors, rapid-fire, profanity)
• Intent Classification: Credential theft, privesc, persistence, lateral movement, crypto theft
• Attack Phase Tracking: Recon → Exploitation → Persistence → Exfiltration

---

Legal Disclaimer

This tool is designed for authorized security research and education only. Deploy only on networks you own or have explicit permission to monitor. Always comply with local laws regarding network deception and monitoring.