Abstract

Wireless Sensor Networks (WSNs) are highly vulnerable to security threats due to their decentralized nature, constrained resources, and open communication channels. Traditional intrusion detection and prevention systems (IDPS) often struggle to provide real-time protection while maintaining network efficiency. The increasing complexity of cyberattacks necessitates advanced techniques for threat mitigation. A major challenge in WSN security is the detection of sophisticated intrusions with high accuracy while minimizing false positives and computational overhead. Conventional rule-based and anomaly-based detection methods exhibit limitations in identifying emerging threats due to their reliance on predefined signatures and static models. Addressing these gaps, a hybrid deep learning-based IDPS is proposed, integrating Convolutional Neural Networks (CNNs) for feature extraction and Long Short-Term Memory (LSTM) networks for sequential pattern learning. The hybrid model is trained on a benchmark WSN intrusion dataset and optimized using the Adam optimizer to enhance detection performance. Experimental evaluation shows that the proposed model achieves an intrusion detection accuracy of 98.6%, significantly outperforming traditional machine learning approaches such as Support Vector Machines (SVM) (91.2%) and Random Forest (94.8%). The system also reduces false positive rates to 1.8%, ensuring reliable threat identification. Moreover, real- time implementation exhibits an average detection latency of 0.35 seconds, making it suitable for resource-constrained WSN environments. These results indicate that the hybrid CNN-LSTM model effectively enhances the security of WSNs, providing a robust defense against evolving cyber threats.

Authors

V. Balajishanmugam¹, A. Christopher Paul², B. Thirunavukarasu³
PPG Institute of Technology, India¹, Karpagam Institute of Technology, India², Queensland University of Technology, Australia³

Keywords

Intrusion Detection, Wireless Sensor Networks, Deep Learning, Cybersecurity, Threat Mitigation