A Multi-Metric Trust-based Approach for Detecting Black Hole Attacks in Wireless Sensor Networks using K-Means Clustering
Main Article Content
Abstract
In Wireless Sensor Networks (WSNs), securing communication against threats like blackhole attacks is essential for maintaining the integrity of the network. This paper proposes a system where the Cluster Head (CH) monitors and evaluates node behavior using manifold metrics: Packet Forwarding Behavior (PFB), Acknowledgment Ratio (AR), Reputation Score (RS), Cooperation Ratio (CR), and Energy Deviation (ED). These metrics help detect abnormal activities, especially packet dropping associated with blackhole attacks. The PFB measures how effectively a node forwards packets, while AR gauges the node's reliability in sending acknowledgments for forwarded packets. The RS is a long-term metric combining AR and PFB to track a node's trustworthiness over time. CR assesses a node's cooperative behavior with neighboring nodes, and ED evaluates energy usage, identifying anomalies in power consumption that could signal malicious activity. To classify nodes as either normal or potentially malicious (blackhole), a K-means clustering algorithm is employed. Nodes are grouped based on the five metrics into two clusters: one for normal nodes and another for suspicious nodes. The algorithm iteratively adjusts the cluster centroids using Euclidean distance until stable clusters are formed or a maximum number of iterations is reached. By applying this approach, the system effectively differentiates between normal and blackhole nodes, improving the security and resilience of WSNs against attacks. A proof of mathematical has proven the applicability of the proposed model. The simulation results shows better result compare with other existing models in terms of performance metrics. The proposed model has effectively detect the balckhole nodes compare with other models.