SmartWSN-IDS: A Hybrid Deep Reservoir and Optimized Tree Model for Routing Attack Detection

Wireless Sensor Networks (WSNs) comprise a large number of low-power sensor nodes deployed in applications such as environmental monitoring, healthcare, military surveillance, and smart cities. Recent studies indicate that over 60% of WSN deployments operate in unattended and hostile environments, making them highly susceptible to security threats. This vulnerability has increased the impact of routing attacks, including Time Division Multiple Access (TDMA) attacks, black-hole attacks, flooding attacks, and gray-hole attacks. These attacks significantly degrade network performance by causing packet loss, excessive energy consumption, and delays in communication. Conventional intrusion detection systems (IDS) for WSNs, which rely on manual rules and predefined thresholds, are inadequate for dynamic and large-scale environments. Such approaches depend heavily on expert knowledge and static monitoring mechanisms, resulting in high false alarm rates, poor scalability, delayed detection, and limited adaptability to emerging or hybrid attacks. To address these limitations, this study proposes a Dual Tree Optimizer-based WSN Routing Attack Detection (DTOWSN-RAD) framework. The framework begins with the utilization of a WSN routing attack dataset, followed by comprehensive data preprocessing steps including normalization, noise removal, and feature selection to improve data quality. For performance benchmarking, existing classification models such as Decision Tree Classifier (DTC), Ridge Classifier (RC), and Linear Discriminant Analysis (LDA) are implemented. The proposed approach integrates Echo State Network (ESN)-based feature extraction with Decision Tree Cost Complexity Pruning (DTCCP) to effectively capture both temporal and structural routing behaviors. This hybrid model enables accurate detection and classification of TDMA, black-hole, flooding, and gray-hole attacks. Experimental results demonstrate that the DTOWSN-RAD framework achieves higher detection accuracy, improved robustness, and better adaptability compared to conventional methods.