Abstract
Nowadays, many fields, such as healthcare, farming, factories, transportation, cities, and homes are connected via network devices. These systems are configured in open environments and are prone to malicious attacks. It is important to protect these systems from intruders and cyberattacks. Due to the increase in data, the diverse nature of devices, and the types of attacks, standard security systems find it difficult to tackle these attacks. Many researchers have worked to address the problem of intrusion detection in networks. Machine learning and deep learning have also been used. Despite the strong literature, the accuracy of the methods is still an open issue. This article presents a model for intrusion detection with improved accuracy using the UNSW-NB15 dataset. The model uses a wrapper-based feature selection technique using machine learning algorithms to select the best features, which are then combined and fed into a Multi-Head Attention-based transformer for getting the predictions. Accuracy, Precision, recall, and F-1 score are used to evaluate the model. The proposed model improves the accuracy of intrusion detection by selecting the most relevant features while reducing the feature space. The results are compared with other state-of-the-art methods to prove the validity of the proposed model.