sigcross: Signal Desktop + CRoSS Steganography

A privacy-focused fork of Signal Desktop that integrates the NeurIPS 2023 "CRoSS" framework. This allows users to securely hide secret images inside standard "container" images using Stable Diffusion.

Prerequisites & Setup

Before running the application, you must set up the environment for the CRoSS steganography feature.

1. Python Environment

You need Python installed on your system. Install the required dependencies:

pip install -r CRoSS/requirements.txt

2. Model Weights (Crucial)

You must download the Stable Diffusion v1.5 weights into the CRoSS/sd15/ directory.

Note: This requires git-lfs. If you don't have it, install it first (e.g., git lfs install).

Run the following commands from the repository root:

# Initialize Git LFS
git lfs install

# Clone the model weights into the specific directory
git clone https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5 CRoSS/sd15

Ensure the folder structure looks like this: sigcross/CRoSS/sd15/ (containing model_index.json, unet/, vae/, etc.)

Performance Note

This feature runs a Stable Diffusion model locally on your machine.

• GPU Recommended: A dedicated GPU (NVIDIA or Apple Silicon) is highly recommended.
• CPU Warning: Running on CPU will be significantly slower.

Usage Guide

To Hide (Encrypt)

1. Attach an image in a conversation.
2. A dialog will appear asking if you want to encrypt the image.
3. Click "Yes, Encrypt".
4. Enter your "Secret Message (Private Key)" (the hidden text/concept) and a "Public Key" (the visible concept).
   • This process uses CRoSS/demo.py.

To Reveal (Decrypt)

1. Right-click on a received image message in the timeline.
2. Select "Decrypt with CRoSS".
3. Enter the Secret Message and Public Key used during encryption.
4. The revealed image will be saved to your Downloads/SignalDecrypted folder.
   • This process uses CRoSS/reveal.py.

Development

Standard Signal build commands apply (using pnpm in this fork):

# Install dependencies
pnpm install

# Generate necessary protobufs and artifacts
pnpm run generate

# Start the application
pnpm start

About the Project

This project was developed by Darpan Bhattacharya for the course DA300 (Term Project) in the MSc in Big Data Analytics program at RKMVERI.

References

• CRoSS Framework: CRoSS: Diffusion Model Makes Controllable, Robust and Secure Image Steganography (NeurIPS 2023).
• Upstream Repository: Signal-Desktop

License & Attribution

This project retains the AGPL-3.0 license from the upstream Signal Desktop repository.