Intrusion detection using decision tree classifier with feature reduction technique

The number of internet users and network services is increasing rapidly in the recent decade gradually. A Large volume of data is produced and transmitted over the network. Number of security threats to the network has also been increased. Although there are many machine learning approaches and methods are used in intrusion detection systems to detect the attacks, but generally they are not efficient for large datasets and real time detection. Machine learning classifiers using all features of datasets minimized the accuracy of detection for classifier. A reduced feature selection technique that selects the most relevant features to detect the attack with ML approach has been used to obtain higher accuracy. In this paper, we used recursive feature elimination technique and selected more relevant features with machine learning approaches for big data to meet the challenge of detecting the attack. We applied this technique and classifier to NSL KDD dataset. Results showed that selecting all features for detection can maximize the complexity in the context of large data and performance of classifier can be increased by feature selection best in terms of efficiency and accuracy.