TuskNet

Your Data, Your Control
Collective Intelligence, Powered by Walrus.

This project won the 1000$ prize in the WALRUS track

Say Hi to TuskNet! A decentralized platform built on the Walrus network, designed to enable secure and privacy-preserving model training across distributed datasets. It ensures that sensitive data remains confidential by leveraging advanced cryptographic methods, including Pedersen commitments, to safeguard model parameters (weights and biases) throughout the training process.

Read our wonderful docs here.

Key Features and Architecture

1. BlockChain Coordination (Walrus)

   TuskNet leverages Walrus' power to manage node registration, emit real-time events, and securely aggregate encrypted data—all seamlessly and without compromise. It ensures every contributor can participate in the process, confident that their efforts are secure and valued.

2. Privacy-Preserving Model Training

   We live in a world where data privacy is no longer optional—it is a necessity. TuskNet embraces this by employing Pedersen commitments, a cutting-edge cryptographic method that ensures your data and model parameters remain confidential.

3. Decentralized Node Participation

   In the past, access to advanced AI development was reserved for the few—those with vast resources and centralized control. TuskNet flips this paradigm on its head. Now, anyone with data can participate as a node in a trustless ecosystem.

4. Encrypted Aggregation

   With TuskNet, nodes submit encrypted model updates—weights and biases—that are aggregated on-chain by smart contracts on the Walrus network. The result? A seamless integration of contributions, enabling better AI models without ever exposing private data.

Run locally

1. First clone the repository.

2. Setup the local walrus server.

   cd Server/server
   npm i 
   node walrus.js

3. Setup the walrus client.

   cd clientnode
   npm i 
   node model.js

4. Initialize the interface.

    pnpm i
    pnpm dev