Clone-Node Detection in Wireless Sensor Networks Using Behavioral Analysis and Machine Learning

Wireless Sensor Networks (WSNs) remain vulnerable to clone-node attacks, in which malicious replicas use the identity or credentials of legitimate sensor nodes. This study presents an intelligent learning-based approach for clone-node detection using behavioral characteristics related to communication, topology, traffic, distance, and energy consumption. Since the WSN-DS dataset does not contain a native clone-node class, a controlled synthetic clone-node scenario was developed from legitimate sensor observations by preserving node identities while introducing behavioral changes associated with replication. A group-aware evaluation strategy was used to evaluate Logistic Regression, Random Forest, 1D-CNN, and a correlation-enhanced DCNN under the same experimental conditions. The results showed different detection characteristics among the evaluated models. Random Forest achieved the strongest overall classification performance, while Logistic Regression also provided strong discrimination with a low false-positive rate. The deep-learning models achieved high clone-detection recall, with the correlation-enhanced DCNN reaching 99.46% recall. These results demonstrate that different learning approaches provide different advantages for clone-node detection and highlight the importance of considering accuracy, detection sensitivity, and false-positive behavior together. The findings provide a useful basis for developing more reliable and adaptive intelligent clone-node detection methods for resource-constrained WSN environments.