Anomaly Detection in IoT Networks Using Differential Evolution and XGBoost

Internet of Things (IoT) is a valuable area that improves the standard of living that cannot be achieved with the current traditional paradigm. With IoT, attack opportunities are also growing. The intrusion detection system effectively detects whether the attack is normal or not. Consequently, classification techniques can be applied to prediction. The concepts of machine learning (ML) and deep learning (DL) in artificial intelligence (AI) technology, which have a greater impact on data science, have led to significant developments in IoT applications. In this paper, anomaly detection in IoT Networks using Differential Evolution and XGBoost (INDEXGB) is presented. In addition, we have also used dimensionality reduction to solve the unbalanced IoTID20 dataset. By determining the fewest features necessary to maintain system performance, the number of features is decreased. The fundamental concept is to use Differential Evolution to select a few features from the IoTID20 datasets and then to use Extreme Gradient Boosting (XGBoost) Learning to compute their accuracy. The process of Differential Evolution (DE) is continued until the smallest set of characteristics with good accuracy is obtained. DE is chosen for its effectiveness in solving global optimization problems, robustness in handling complex and noisy objective functions, simplicity in implementation, and versatility across different types of optimization. The proposed INDEXGB model reduced training and testing time while still achieving an accuracy of 83.72%.