AN INTELLIGENT ENSEMBLE-BASED SYSTEM THAT USES THE NETWORK INTRUSION DETECTION SYSTEM PARADIGM TO IDENTIFY AND FIGHT INTRUSIONS IN COMPUTER NETWORKS

In recent years, the Internet of Things (IoT) has undergone tremendous change, allowing people to automate mundane, everyday activities. Connecting several types of physical devices with distinct functions allows for this to be accomplished. In order to enhance intrusion detection systems, machine learning has emerged as the crucial option. As the Internet has grown in popularity and the number of suspicious activities or intrusions has accelerated, research into network intrusion detection systems (NIDS) has emerged as a pressing concern in the field of information and network security. By using a classification method, intrusion detection systems (IDS) can distinguish between normal and abnormal incoming network traffic, which is represented as a feature vector. This helps in the detection of intrusions that violate a computer network’s security policies and mechanisms and compromise CIA (confidentiality, integrity, and availability). It has been noted that classification performance is negatively impacted by feature vectors with large dimensionality in practice. A novel hybrid feature selection strategy was developed to lower the dimensionality while maintaining performance. Its efficacy was evaluated on the KDD Cup’99 dataset using the classifiers Naive Bayes and C4.5. Based on the aforementioned dataset and classifiers, two sets of experiments were carried out using the full feature set and reduced feature sets obtained using four popular feature selection methods: Correlation-based Feature Selection (CFS), Consistency-based Feature Selection (CON), Information Gain (IG), Gain Ratio (GR), and the proposed method. Classifier Naıve Bayes achieved a classification accuracy of 97.5% in the first trial, whereas C4.5 achieved 99.8%. Using the IG approach, the classifiers’ greatest performance (accuracy) was 99.1 and 99.8 percent in the second set of testing.