FL-DABE-BC: A Privacy-Enhanced Decentralized Authentication and Secure Communication Framework for FL in IoT-Enabled Smart Cities

Abstract

Federated Learning (FL) offers a distributed approach to machine learning that preserves data privacy by avoiding the exchange of sensitive IoT sensor information. This paper introduces a novel IoT framework that integrates advanced security tools to tackle key privacy and security challenges. It employs Decentralized Attribute-Based Encryption (DABE) for decentralized authentication and data encryption, Homomorphic Encryption (HE) for secure computations on encrypted data, Secure Multi-Party Computation (SMPC) for collaborative processing, and Blockchain for distributed ledger management and transparent communication. In this system, IoT devices encrypt data locally with DABE, while initial model training occurs on cloud servers within an immutable blockchain network that supports peer-to-peer authentication. Encrypted model weights are then transferred to the fog layer via HE and aggregated using SMPC, after which the FL server updates and distributes the global model to the IoT devices. This innovative framework effectively addresses the challenges of secure decentralized learning, enabling privacy-preserving, efficient, and secure federated learning for IoT applications and real-time analytics. © 2025 Copyright held by the owner/author(s).