Lightweight Cryptography-Enhanced Federated Learning for Secure IoT Environments
Year: 2026
Type: conference-paper
Abstract: Federated learning (FL) enables collaborative model training in Internet of Things (IoT) environments while preserving data locality; however, transmitted model updates remain vulnerable to eavesdropping and in-transit tampering over insecure communication channels. This paper presents a hardware-level evaluation of lightweight cryptographic protection for FL update exchange on resource-constrained ESP32 devices. In the proposed framework, client-side model updates are protected during transmission using the lightweight block ciphers SIMON, SPECK, and PRESENT with integrity verification. The framework was evaluated on the CICIoT2023 dataset under a practical FL setting with three clients and five communication rounds. The main focus of the study is the analysis of securityinduced overhead in terms of RAM/ROM footprint, encryption and decryption latency, throughput, and energy consumption. In addition, the impact of protected update exchange on FL model performance was assessed. The results show that SPECK provides the most favorable efficiency on ESP32, achieving an encryption latency of$0.1348 \mu \mathrm{s} / \mathrm{B}$and energy consumption of 138.3 mJ, while the protected federated model achieved 89.12% accuracy and outperformed the centralized baseline of 86.15 % under the evaluated conditions. These findings demonstrate that lightweight confidentiality and integrity protection can be integrated into FL update exchange on microcontroller-class IoT devices with practical overhead.
Institutions: Satbayev University, University of the West of England
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7168251069