Security-Research-Toolkit

Security Research Toolkit — Video and image analysis tool for neural inpainting and AI-generated content detection with SORA signature extraction, temporal consistency analysis, CNN artifact detection, CPU/CUDA device selection, multi-format support, and colorama-styled terminal interface

 __          _______ _    _  _____    _____                      _ _          
 \ \        / / ____| |  | |/ ____|  / ____|                    (_) |         
  \ \  /\  / / (___ | |  | | (___   | (___   ___  ___ _   _ _ __ _| |_ _   _  
   \ \/  \/ / \___ \| |  | |\___ \   \___ \ / _ \/ __| | | | '__| | __| | | | 
    \  /\  /  ____) | |__| |____) |  ____) |  __/ (__| |_| | |  | | |_| |_| | 
     \/  \/  |_____/ \____/|_____/  |_____/ \___|\___|\__,_|_|  |_|\__|\__, | 
                                                                        __/ | 
 _____                               _       _______          _ _    _|___/  
|  __ \                             | |     |__   __|        | | |  (_) |    
| |__) |___  ___  ___  __ _ _ __ ___| |__      | | ___   ___ | | | ___| |_   
|  _  // _ \/ __|/ _ \/ _` | '__/ __| '_ \     | |/ _ \ / _ \| | |/ / | __|  
| | \ \  __/\__ \  __/ (_| | | | (__| | | |    | | (_) | (_) | |   <| | |_   
|_|  \_\___||___/\___|\__,_|_|  \___|_| |_|    |_|\___/ \___/|_|_|\_\_|\__|  

Security Research Toolkit

Video & image analysis for neural inpainting and AI-generated content detection

Features • Getting Started • Configuration • Usage • Project Structure • FAQ

---

Official Links

Resource	URL
Repository	https://github.com/timanmoh/Security-Research-Toolkit
Issues	https://github.com/timanmoh/Security-Research-Toolkit/issues
OpenAI Sora	https://openai.com/index/sora
Reality Defender (Sora detection)	https://www.realitydefender.com/insights/detecting-sora-videos

---

Features

Analysis Video frame validation (MP4, AVI, MKV, MOV, WebM, WMV, FLV) Image validation (PNG, JPG, BMP, WebP, TIFF) Input path validation & format detection Output directory management Detection SORA signature feature extraction Temporal consistency analysis Interpolation detection CNN region/artifact detection

Processing Neural inpainting pipeline (frame/mask) Batch video frame processing CPU/CUDA device selection GPU status via nvidia-smi Interface Interactive terminal interface Colored output (colorama) Multi-language (EN/RU) Settings persistence (JSON)

---

Getting Started

Prerequisites

• Python 3.4 or higher
• Optional: NVIDIA GPU with CUDA for accelerated processing
• Optional: PyTorch with CUDA support (auto-detected)

Installation

git clone https://github.com/timanmoh/Security-Research-Toolkit.git
cd Security-Research-Toolkit
pip install -r requirements.txt
python main.py

Dependencies

Package	Version	Purpose
colorama	≥0.4.6	Cross-platform colored console output

Note: Neural inpainting and detection modules use default implementations. Advanced models (AOT-GAN, LaMa, ProPainter) can be integrated for higher accuracy.

---

Configuration

Settings are stored in settings.json in the project root. Edit via menu Settings or command settings.

Example settings.json:

{
  "device": "auto",
  "output_dir": "./output",
  "log_level": "info",
  "language": "en"
}

Parameter	Values	Description
device	auto, cpu, cuda	Compute device; auto selects GPU if available
output_dir	path	Output directory (relative or absolute); created on first save
log_level	debug, info, warning, error	Console verbosity
language	en, ru	Interface language

Tip: Use device=cpu for headless servers without GPU. Use log_level=debug when troubleshooting.

---

Usage

Interactive Mode (default)

python main.py
# or
python main.py --interactive

Main menu:

┌──────────────────────────────────────────────────────────────┐
│  [1] Install dependencies                                     │
│  [2] Start                                                    │
│  [3] About                                                    │
│  [4] Settings                                                 │
└──────────────────────────────────────────────────────────────┘

  Choose [1]-[4]: 2

Command menu:

> run C:\videos\sample.mp4
> validate ./images/frame_001.png
> gpu
> settings
> help
> quit

CLI Mode

python main.py --run <video_or_image_path>
python main.py --validate <path>
python main.py --gpu

Command	Description
run <path>	Process video or image (detect + inpainting pipeline)
validate <path>	Validate input file and report type (video/image)
gpu	Show NVIDIA GPU status (nvidia-smi)
install / 1	Install dependencies from requirements.txt
about	Show project info from README
settings	View/edit settings (device, output_dir, log_level, language)
help	Show command menu
quit / exit	Exit application

---

Project Structure

Security-Research-Toolkit/
├── main.py                 # Entry point (CLI, interactive menu)
├── settings.json           # User settings (created on first save)
├── requirements.txt        # Python dependencies
├── README.md
│
├── gui/
│   ├── __init__.py
│   └── main_window.py      # Terminal interface (banner, menu, sections)
│
├── core/
│   ├── __init__.py
│   ├── processor.py        # Video/image processing pipeline
│   ├── inpainting.py       # Neural inpainting (region filling)
│   └── validator.py        # Input path validation
│
├── detection/
│   ├── __init__.py
│   ├── detector.py         # CNN region/artifact detection
│   ├── signature.py        # SORA signature analysis
│   └── temporal.py         # Temporal consistency, interpolation
│
└── utils/
    ├── __init__.py
    ├── settings.py         # Load/save settings.json
    ├── file_handler.py     # File operations, path sanitization
    ├── gpu_manager.py      # GPU info, device suggestion
    └── logger.py           # Logging utilities

---

FAQ

What video formats are supported?

MP4, AVI, MKV, MOV, WebM, WMV, and FLV. Image formats: PNG, JPG, JPEG, BMP, WebP, TIFF, TIF.

Does it work without a GPU?

Yes. The toolkit runs on CPU by default. Set device=cpu in settings or use auto to let the app detect CUDA availability. GPU acceleration is optional.

What is SORA signature detection?

SORA is OpenAI's text-to-video model. The detection module analyzes temporal consistency, generation artifacts, and feature signatures to identify AI-generated content. The current implementation provides extensible interfaces for integrating custom classifiers.

How do I add real neural inpainting?

Extend the base implementation in core/inpainting.py with a model such as AOT-GAN, LaMa, or ProPainter. The inpaint_frame(frame_region, mask_region, device) function expects numpy/tensor input and returns the inpainted region.

Settings are not saving. What to check?

Ensure the project directory is writable. settings.json is created in the project root. On Windows, run as administrator if the folder has restricted permissions. Check that only valid keys (device, output_dir, log_level, language) and values are used.

nvidia-smi not found / GPU not detected

Install NVIDIA drivers and ensure nvidia-smi is in PATH. On Linux, it is typically in /usr/bin. The toolkit falls back to CPU if no GPU is detected. PyTorch CUDA support is optional and checked at runtime.

Can I use this for production?

This is a research/educational toolkit. Detection and inpainting modules are extensible base implementations. For production use, integrate validated models, add error handling, and perform security audits. See the Disclaimer below.

---

Disclaimer

This project is intended exclusively for educational and security research purposes. Use it only on content you own or have explicit permission to analyze. Do not use it to create, distribute, or analyze deepfakes or manipulated media for deceptive purposes. The authors are not responsible for misuse. AI-generated content detection is an evolving field; results may be inaccurate. Always comply with local laws and platform terms of service.

---

If this project helped your research, consider giving it a star.

ETH: 0x6f1A3c5E9B2a4D6e8C0b3F5a7D9c1E3b5A7f28e1