Federated Learning System for IoT Devices (Anomaly Detection)

#fedavg #kafka #apache

Good Morning Y'ALL 🙌 . Wanted to share a university group project we worked on:
Federated Learning System for IoT Devices (Anomaly Detection)
_https://github.com/imadbenmadi/FLED _

We built a distributed (multi-broker) federated learning system where IoT edge devices train models locally **and share only model updates to the **FedAvg Server.
FedAvg Server Create and Update, and Save the Global Models. preserving data privacy and avoiding the limitations of centralized training.

We simulated 2,401 IoT devices streaming sensor data from a dataset. Data ingestion was handled with Kafka, per-device stream processing and local modeling with Flink, and evaluation with Spark.

Local and global models, along with evaluation results, were stored in TimescaleDB, and Grafana was used for real-time dashboards and anomaly alerts.