Enhancing Data Reliability and Synchronization in Wireless Sensor Networks using RAE-DRNet with Federated Learning

Abstract

Wireless Sensor Networks (WSNs) play a critical role in modern Internet of Things (IoT) applications, enabling distributed sensing and real-time decision-making; however, their performance is often degraded by noisy data, communication delays, and synchronization issues. This study aims to address these challenges by proposing a novel Reconstruction-Assisted Enhancement-Data Reliability Network (RAE-DRNet) integrated with federated learning to improve data reliability, synchronization, and overall network efficiency. The proposed methodology utilizes autoencoder-based reconstruction to recover corrupted and missing sensor data, combined with residual learning and attention mechanisms to enhance feature representation. Additionally, a delay-aware and buffer-assisted federated learning framework is incorporated to handle asynchronous updates and reduce communication overhead. The model is evaluated using the WSN Multi-Hop dataset, which captures realistic network conditions including latency and multi-hop communication behavior. Experimental results demonstrate that the proposed RAE-DRNet significantly outperforms existing approaches, achieving an accuracy of 94.8%, latency of 75 ms, energy consumption of 0.68 J, and synchronization error of 5.3%. These improvements highlight the effectiveness of the model in enhancing data quality and reducing network inefficiencies. In conclusion, the proposed approach provides a scalable, efficient, and reliable solution for federated learning-based WSN environments, with strong potential for real-world deployment in dynamic and resource-constrained IoT systems.