Optimized Federated Learning for Trustworthy Edge Decision-Making in IoT Consumer Electronics
Abdul Rehman, Mahmood ul Hassan, Khalid Mahmood, Kamran Ahmad Awan, Niyaz Ahmad Wani, Muhammad Shahid Anwar · IEEE Transactions on Consumer Electronics · 2025
The Internet of Things (IoT) is fundamentally transforming industries by enabling vast networks of interconnected devices to support decision-making processes at the edge of the network. Addressing the critical challenge of digital trust in IoT-enabled AI systems, this paper presents an Optimized Trust-Oriented Federated Learning Framework for IoT (TOFL-IoT) designed for decision-making in consumer electronics. TOFL-IoT integrates federated learning with a multi-level trust evaluation mechanism to enhance model reliability, scalability, and efficiency. The simulations consist of 100 IoT devices and 10 edge servers, TOFL-IoT achieved model accuracy of up to 92.2% under diverse data quality conditions.. The framework demonstrated resilience with an accuracy of 84.5% under high device dropout rates and 87.6% under adversarial attacks, consistently surpassing comparable methods by 3-7%. Additionally, TOFL-IoT reduced communication overhead by 20% relative to baseline approaches. Privacy preservation was also robust, with privacy scores ranging from 0.76 to 0.89 across varying scenarios.