Hybrid LSTM-Random Forest for Intrusion Detection in Wireless Sensor Networks: Enhanced DoS Classification with Explainable AI

Abstract

Wireless Sensor Networks (WSN) have emerged as one of the most active study topics in computer science due to their vast range of applications, which include crucial military and civilian uses. To ensure the security and dependability of WSN services, an Intrusion Detection System (IDS) should be implemented. This IDS must be compatible with the features of WSNs and capable of identifying the greatest number of security risks. Using the WSN dataset, this research proposes a new hybrid model that combines LSTM and Random Forest to help detect and categorize four forms of Denial of Service (DoS) attacks: blackhole, grayhole, flooding, and scheduling. The "proposed model" surpasses LSTM, GRU, RNN, CNN, CNN-LSTM, Random Forest, GaussianNB, and Decision Tree in attack detection, as indicated by the highest accuracy, precision, recall, F1-Score, and AUC accuracy score of 0.996, 0.98, 0.98, 0.98, and 0.99, respectively. By offering insights into the decision-making process and facilitating a better comprehension of the feature contributions to attack detection, the application of Explainable Artificial Intelligence (XAI) approaches to the Random Forest model analysis enhanced the interpretability of the results.