Enhancing Security Features in WSNs using Autoencoder-Based Intrusion Detection and ECC with Dynamic Key Rotation

Objectives: To propose an intelligent, dual-layered security model to detect anomalies in wireless sensor networks for secured data transmission in large-scale WSN environments. The proposed model integrates deep learning and adaptive encryption methods to address security breaches in smart environments due to their lightweight and distributed nature. Methods: Auto-Encoder Neural Network (AEN) is employed to detect abnormal behavior in sensor nodes based on reconstruction error thresholds. Post validation, Elliptic Curve Cryptography (ECC) is integrated with Dynamic Key Rotation (DKR) to encrypt the data packets. It adjusts the cryptographic keys based on network conditions to enhance the node resistance to replay and key compromise attacks. Multi-hop swarm route optimization is used for the finest routing process to ensure the node reaches the destination in the optimal route. NS2 simulation is carried out with three sets of datasets, NSL-KDD, CICIDS, and simulated data with parameters to assess the performance of the proposed AEN-ECC model to ensure real-time attack resilience, end-to-end secure communication and scalability to larger WSNs under dynamic network conditions. AEN-ECC is compared against state-of-the-art machine learning-based security methods like SVM-AES, RF-KSS, and XGBoost-FR. Findings: The proposed intelligent, adaptive, and lightweight AEN-ECC model achieved promising results with a 90.2% Packet Delivery Ratio, 91.4% Confidentiality Rate, 90.6% Attack Detection Accuracy, 90.1% Robustness to Node Failures, 92.7% network life time and 28.5 ms reduced Latency and stands as a solution for WSNs that ensure both confidentiality and anomaly resilience. Novelty: This research moves beyond static security protocols and presents a unified security model that learns from network behavior and adapts dynamic encryption by using an unsupervised AEN enhanced with ECC and DKR to provide scalable and energy-aware model capable of addressing the real-world security challenges faced by WSNs. Keywords: Wireless Sensor Networks, Network Security, Elliptic Curve Cryptography, Dynamic Key Rotation, Intrusion Detection