Energy-Efficient and Secure Anomaly Detection Using Neural Auto Encoder In Wireless Sensor Networks

The study proposes an energy-efficient and secure anomaly detection framework for wireless sensor networks with a lightweight neural auto encoder that enhances precision and accuracy. A five-layer auto encoder with Mean Absolute Error (MAE) reconstruction loss and optimization, input normalization, and anomaly detection using reconstruction error thresholding was implemented. The neural auto encoder was lightweight and used to identify WSN anomalies based on reconstruction error thresholding. The performance was evaluated in terms of false alarm rate, energy consumption, latency, and network lifetime. The presented neural auto encoder had a high accuracy of 98.36%, reduced false alarms, reduced energy consumption, and enhanced detection reliability. The improvement was statistically significant with p = 0.001. The lightweight auto encoder guarantees secure anomaly detection, low power consumption, adaptive thresholding, reduced overhead, extended network lifetime, and improves the overall performance of wireless sensor networks.