DYNAMIC ROLE-BASED SECURITY FRAMEWORK WITH HYBRID VOTING CLASSIFIERS FOR SENSOR NETWORK PROTECTION

Abstract

Wireless Sensor Networks (WSNs) play a crucial role in modern communication systems by enabling efficient data collection and monitoring across applications such as environmental sensing, healthcare, military surveillance, and smart cities. However, the open communication architecture and limited security mechanisms of WSNs make them highly vulnerable to routing attacks, including flooding, TDMA manipulation, gray hole, and black hole attacks. These threats disrupt network communication, increase energy consumption, and reduce overall system reliability and lifespan. Traditional intrusion detection systems in WSNs rely on rule-based and signature-based techniques that detect known attack patterns. While effective for predefined threats, these methods struggle to identify novel and complex attacks in dynamic network environments. To address these limitations, this study proposes a machine learning–based routing attack detection framework that analyzes network communication parameters to identify malicious behavior. The framework incorporates baseline classifiers such as Decision Tree Classifier (DTC), Ridge Classifier (RC), and Linear Discriminant Analysis (LDA) to evaluate detection performance. Additionally, a novel hybrid model, termed Deep Reservoir Routing Defense (DRRD), is introduced by integrating Echo State Network (ESN) with Decision Tree Cost Complexity Pruning (DTCCP). The ESN performs reservoir-based feature transformation to capture dynamic communication patterns, while DTCCP ensures accurate classification based on transformed features. To enhance model performance, preprocessing techniques including label encoding, missing value handling, and SMOTE-based data balancing are applied. Experimental results demonstrate that the proposed DRRD model significantly outperforms traditional methods in detection accuracy, offering a scalable and efficient solution for securing WSN environments