Federated Learning in IoT

Python • Flower
Dec 2024 – Present

A federated learning system designed for Internet of Things (IoT) environments using the Flower framework. This project enhances model accuracy, reduces latency, and ensures secure aggregation across IoT nodes.

Showcase

• 30%+ Improvement in Model Accuracy over traditional centralized approaches, demonstrated through experimental evaluations across heterogeneous IoT data sources.
• Real-world Simulation Using Raspberry Pi: Configured a centralized server on a laptop to orchestrate distributed training over IoT devices; deployed a Raspberry Pi client to process 10,000+ sensor data points per training cycle.
• Communication Optimization: Engineered and optimized protocols to achieve a 25% reduction in latency, enabling secure real-time model synchronization across nodes.

Key Features

• Federated Learning with Flower: Enables decentralized model training directly on IoT endpoints, facilitating privacy-preserving intelligence.
• Flexible Architecture:
  • app.py: Main entry point.
  • server.py: Federated learning server coordination.
  • temperature.py: Illustrative sensor data client.
  • models/: Model definitions.
  • templates/: (If applicable) UI or configuration templates.
• Sample Data Included: dht_readings.csv, cleaned_dht_readings.csv, and data.csv for testing and validation workflows.
• Support Scripts: fake.py, new1.py, and example/ for generating and experimenting with sample workflows.
• Scalable and Modular: Easily extendable to new model architectures, IoT devices, and aggregation strategies.

---

Usage Guide

1. Prerequisites

• Python 3.x installed
• Install dependencies:

  pip install -r requirements.txt