PoCSmith

AI-Powered Proof-of-Concept Generator for Security Research

PoCSmith is an AI model fine-tuned on exploit code and CVE data to assist security researchers in generating proof-of-concept exploits and shellcode for defensive purposes.

Author: Regaan

Ethical Use Only

This tool is designed exclusively for defensive security research and authorized testing:

• Penetration testing on systems you own or have permission to test
• Security research and vulnerability analysis
• Educational purposes in controlled environments
• NOT for malicious attacks or unauthorized access

Use responsibly. Follow all applicable laws and regulations.

Features

• AI-Powered Generation - CodeLlama-7B fine-tuned on 1,472 exploit samples
• CVE Integration - Fetch vulnerability data from NVD API
• Multi-Platform Shellcode - x86, x64, ARM support
• Simple CLI - Easy command-line interface
• High Quality - 78.4% token accuracy

Quick Start

Installation

git clone https://github.com/noobforanonymous/PoCSmith.git
cd PoCSmith

python3 -m venv venv
source venv/bin/activate

pip install -e .

Usage Examples

# Generate exploit from CVE
python src/cli/main.py cve CVE-2024-1234

# Generate shellcode
python src/cli/main.py shellcode --platform linux_x64 --type reverse_shell --lhost 10.10.14.5 --lport 4444

# Generate from vulnerability description
python src/cli/main.py generate --vuln "buffer overflow" --target "Apache 2.4"

# List available options
python src/cli/main.py list-platforms
python src/cli/main.py list-payloads

Model Details

• Base Model: CodeLlama-7B
• Training: QLoRA 4-bit quantization
• Dataset: 1,472 samples (CVE-Exploit pairs + shellcode)
• Performance: 78.4% token accuracy, 30% loss reduction
• Training Time: 3h 17min on RTX 4050 (6GB VRAM)

Project Structure

PoCSmith/
├── src/
│   ├── parsers/          # CVE parsing
│   ├── generators/       # Exploit & shellcode generation
│   ├── formatters/       # Output formatting
│   ├── cli/              # Command-line interface
│   └── core/             # Configuration
├── models/
│   └── pocsmith-v1/      # Fine-tuned AI model (LoRA adapters)
├── data/                 # Training data
├── docs/                 # Documentation
└── tests/                # Unit tests

Documentation

• Usage Guide
• Setup Instructions
• Fine-tuning Details
• System Architecture

Requirements

• Python 3.11+
• CUDA-capable GPU (6GB+ VRAM recommended)
• 20GB disk space

Dependencies

torch>=2.0.0
transformers>=4.35.0
peft>=0.7.0
bitsandbytes>=0.41.0
click>=8.1.0

Example Output

Shellcode Generation

$ python src/cli/main.py shellcode --platform linux_x86 --type reverse_shell --lhost 10.10.14.5 --lport 4444

PoCSmith v1.0

[*] Generating reverse_shell for linux_x86...
Loading PoCSmith model...
Model ready!

/*
 * Shellcode for Linux/x86
 * - Calls socket() -> connect() -> dup2() -> execve()
 * - Tested on Ubuntu, Debian
 * - Length: 160 bytes
 */

Contributing

Contributions are welcome. Please fork the repository, create a feature branch, and submit a pull request.

License

MIT License - See LICENSE file

Disclaimer

FOR EDUCATIONAL AND DEFENSIVE SECURITY RESEARCH ONLY

I am not responsible for misuse of this tool. Users must obtain proper authorization before testing, follow responsible disclosure practices, and comply with all applicable laws.

Acknowledgments

• CodeLlama (Meta AI)
• NVD (NIST)
• Exploit-DB
• Metasploit Framework
• Hugging Face

Built for the security research community.

Version 1.0
By Regaan