NULLSEC AI v3.0 - Offline-First AI Guide

🚀 Revolutionary Features

NO API KEYS REQUIRED - NULLSEC AI v3.0 works 100% offline with no accounts, no subscriptions, and no API keys needed!

✨ What's New in v3.0

1. Offline-First Architecture

Works without any internet connection
No API keys or user accounts required
Privacy-focused - all data stays local
Rule-based expert system as fallback

2. Multiple Free AI Models

All models are open source and run locally:

Model	Size	Specialty	Best For
DeepSeek Coder	6.7B	Code generation	Exploit development, shellcode
CodeLlama	13B	Meta's code model	General coding, payloads
WizardCoder	15B	Enhanced coding	Complex exploit chains
Mistral	7B	Fast general	Quick analysis, recon
Mixtral	8x7B	Expert mixture	Advanced scenarios
OpenHermes	7B	Instruction tuned	Detailed explanations
Solar	10.7B	Advanced reasoning	Attack planning
Phi-2	2.7B	Efficient	Low resource systems
Orca2	13B	Reasoning	Complex problem solving
Neural Chat	7B	Conversational	Interactive pentesting

3. Built-in Expert System

When no AI is available, the system uses:

100+ pre-programmed pentesting rules
Methodology-based command generation
Tool recommendation engine
Best practice guidance

4. Supported AI Providers

Ollama (Recommended) - Full featured local AI
LM Studio - GUI-based local AI
GPT4All - Easy to use desktop app
LocalAI - OpenAI-compatible local server

📦 Quick Start (3 Methods)

Method 1: Ollama (Recommended - Best Performance)

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull your first model (choose one or all)
ollama pull deepseek-coder:6.7b    # Best for exploits
ollama pull codellama:13b          # General purpose
ollama pull mistral:7b             # Fast and efficient

# Launch NULLSEC AI
cd /home/antics/nullsec
python3 nullsec-ai.py

Method 2: No AI Setup Required

# Just run it - works immediately with expert system
cd /home/antics/nullsec
python3 nullsec-ai.py

# The system will use rule-based expert mode
# Still provides pentesting commands and advice

Method 3: LM Studio (GUI Option)

Download LM Studio from https://lmstudio.ai
Download a model through the GUI
Start the local server (default port 1234)
Run NULLSEC AI - auto-detects LM Studio

💡 Usage Examples

Interactive Mode

$ python3 nullsec-ai.py

# Set your target
set target 192.168.1.100

# Choose attack category
set category web

# Ask questions
How do I test for SQL injection?
What ports should I scan first?
Generate a reverse shell payload for Linux

# Execute commands directly
execute nmap -sV 192.168.1.100

# View available models
models

# Install new model
install wizardcoder:15b

Example Session

nullsec-ai(general) > set target testsite.com
[+] Target set to: testsite.com

nullsec-ai(general) > set category web
[+] Category set to: web

nullsec-ai(web)[testsite.com] > enumerate this web application

[AI Response]

WEB APPLICATION ENUMERATION:

1. Technology Detection:
   whatweb http://testsite.com
   wappalyzer testsite.com

2. Directory/File Discovery:
   gobuster dir -u http://testsite.com -w /usr/share/wordlists/dirb/common.txt
   ffuf -u http://testsite.com/FUZZ -w wordlist.txt

3. Subdomain Enumeration:
   subfinder -d testsite.com
   amass enum -d testsite.com

4. Parameter Discovery:
   arjun -u http://testsite.com

5. Vulnerability Scanning:
   nikto -h http://testsite.com
   nuclei -u http://testsite.com

nullsec-ai(web)[testsite.com] > execute gobuster dir -u http://testsite.com -w /usr/share/wordlists/dirb/common.txt
[>] gobuster dir -u http://testsite.com -w /usr/share/wordlists/dirb/common.txt

[gobuster output...]

🎯 Advanced Features

1. Model Management

# List installed models
nullsec-ai> models

# Install specific model
nullsec-ai> install deepseek-coder:6.7b

# Change preferred model (edit ~/.nullsec-ai-v3.json)
{
  "preferred_provider": "ollama",
  "preferred_model": "deepseek-coder:6.7b"
}

2. Command Execution

# Execute commands and log to knowledge base
nullsec-ai> execute nmap -sV 192.168.1.0/24

# View command history
nullsec-ai> history

3. Knowledge Base

The AI learns from your attacks and stores:

Commands executed
Vulnerabilities found
Attack patterns
Successful exploits

Database location: /home/antics/nullsec/.nullsec-ai-v3.db

🔧 Configuration

Edit /home/antics/nullsec/.nullsec-ai-v3.json:

{
  "preferred_provider": "ollama",
  "preferred_model": "deepseek-coder:6.7b",
  "temperature": 0.7,
  "max_tokens": 2000,
  "learning_enabled": true,
  "auto_execute": false
}

Options:

preferred_provider: ollama, lmstudio, gpt4all, localai
preferred_model: Model name to use
temperature: 0.0-1.0 (creativity level)
max_tokens: Response length limit
learning_enabled: Store attack patterns
auto_execute: Auto-run AI-suggested commands (dangerous!)

📊 Model Comparison

Small & Fast (2-7B parameters)

Phi-2 (2.7B) - Very fast, good for low-spec systems
Mistral (7B) - Excellent speed/quality balance
OpenHermes (7B) - Great for conversations
Neural Chat (7B) - Good interaction quality

Use when: Quick responses needed, limited resources

Medium (13B parameters)

CodeLlama (13B) - Meta's code specialist
Orca2 (13B) - Strong reasoning abilities

Use when: Balance of speed and capability

Large & Powerful (15B+ parameters)

WizardCoder (15B) - Enhanced code generation
Mixtral (8x7B = ~47B) - Expert mixture, very capable
Solar (10.7B) - Advanced reasoning

Use when: Complex exploit chains, detailed analysis

Specialized

DeepSeek Coder (6.7B) - RECOMMENDED for pentesting
- Trained specifically on code and security
- Excellent at exploit generation
- Great balance of size and capability

🚀 Performance Tips

Faster Responses

Use smaller models (Phi-2, Mistral 7B)
Reduce max_tokens in config
Use GPU acceleration if available

Better Quality

Use larger models (Mixtral, WizardCoder)
Increase temperature for creativity
Provide detailed context

Resource Management

# Check Ollama status
ollama list

# Monitor resource usage
htop

# Clear GPU cache (if applicable)
ollama stop <model>

🔒 Privacy & Security

100% Private:

All processing happens locally
No data sent to external servers
No telemetry or tracking
No API keys or accounts

Security Note: The AI runs on your machine and can suggest potentially dangerous commands. Always:

Review commands before executing
Use in isolated test environments
Understand what commands do
Keep auto_execute disabled

🆘 Troubleshooting

"No AI providers detected"

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model
ollama pull deepseek-coder:6.7b

# Restart NULLSEC AI
python3 nullsec-ai.py

"Model not found"

# List available models
ollama list

# Pull the specific model
ollama pull <model-name>

Slow Responses

# Use a smaller model
ollama pull mistral:7b

# Then in config, set:
"preferred_model": "mistral:7b"

Out of Memory

# Use Phi-2 (smallest)
ollama pull phi:2.7b

# Or increase swap space
sudo fallocate -l 8G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile

📚 Example Queries

Reconnaissance

"How do I discover all subdomains for example.com?"
"What's the best way to enumerate SMB shares?"
"Generate a comprehensive recon plan for 192.168.1.0/24"

Exploitation

"Create a reverse shell payload for Windows"
"How to test for SQL injection in POST parameters?"
"Generate a XXE attack payload"
"What's the best approach to exploit EternalBlue?"

Post-Exploitation

"How to escalate privileges on Linux?"
"Commands to dump credentials from Windows"
"How to establish persistence on a compromised system?"

Evasion

"How to bypass Windows Defender when running Mimikatz?"
"Obfuscate this PowerShell payload"
"Techniques to avoid IDS detection during port scanning"

🎓 Learning Resources

The AI provides guidance on:

OWASP Top 10 exploitation
Network pentesting methodologies
Privilege escalation techniques
Wireless security attacks
Cloud penetration testing
Mobile app security
Red team operations

⚡ Quick Commands

# Fast start with specific target
echo -e "set target 192.168.1.100\nscan this target\nexit" | python3 nullsec-ai.py

# Get web enumeration commands
echo -e "set category web\nenumerate web app\nexit" | python3 nullsec-ai.py

# Check installed models
ollama list

# Update Ollama
curl -fsSL https://ollama.com/install.sh | sh

🌟 Integration with NULLSEC Framework

The AI integrates seamlessly with NULLSEC:

# From NULLSEC launcher
python3 nullsec-launcher.py
[I] AI Console    # Launches AI assistant

# Direct from command line
python3 nullsec-ai.py

# In scripts
python3 nullsec-ai.py < queries.txt

📈 Future Enhancements

Planned features:

Fine-tuned models specifically for pentesting
Automated exploit chain generation
Integration with Metasploit
Real-time vulnerability database queries
Collaborative multi-agent attacks
Visual attack graph generation

🎉 Summary

NULLSEC AI v3.0 brings professional AI-powered pentesting to everyone:

✅ No API keys or accounts - 100% free and private
✅ Works offline - No internet required
✅ 10+ AI models - Choose based on your needs
✅ Expert fallback - Works even without AI
✅ Learns from you - Builds knowledge base
✅ Easy to use - Simple command interface

Get started in 2 minutes - no signup, no payment, no tracking!

Author: bad-antics development
Repository: github.com/bad-antics/nullsec
Version: 3.0
License: For authorized security testing only

Uh oh!

FilesExpand file tree

NULLSEC_AI_V3_GUIDE.md

Latest commit

History