Nillosphere

Say Hi to CaliFusion! A decentralized platform built on the ICP and Calimero 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.

The Calimero network serves as the backbone for decentralised coordination, facilitating secure participation of nodes in training and enabling on-chain aggregation of encrypted data.

Federated Learning (FL) is a machine learning technique that allows multiple devices or nodes to collaboratively train a model without sharing their local data.

Instead of sending data to a central server, each node processes the data locally and shares only the model updates (like weights and biases). This approach has several advantages, especially when privacy and data security are priorities.