Main Article Content
Intrusion detection system (IDS) can be hardware or software which is used in networking to monitor the network. This is useful when malicious activity is to be managed for network however mounting an IDS is challenging task because attackers always find a different way to attack in the network. Several research has been done on IDS but still there are some issues exists like de An abundance of similar researches had done to related areas which used machine learning process into both host and network-based intrusion systems. But doing the same initially presented few problems like accuracy, high false alarm rate ,low detection rate etc. nowadays due to the very much development in of machine learning which outcome the vast improvements in machine learning algorithms . This paper present a framework model in which many classifiers are used to detect the malicious activity by using machine learning methods. Research uses historical data to apply Intrusion detection on Host machine to detect malicious node. NSL-KDD datasets are used to test the data and evaluation. Based on the NSL-KDD datasets base station is trained so that if any malicious activity is done on the machine then IDS will detect that malicious node based on anomalous behaviour. We define abnormal behaviour by deviating from their normal behaviour. For the training of Host machine Supervised Learning is used which requires already been labelled training data and supervised learning algorithms are used for prediction. Our research also used some data mining algorithms such as KNN (k-nearest neighbour), decision trees, and Ensemble and hybrid algorithms. Hybrid algorithm is the combination of KNN, DT, and Ensemble Algorithm which is more reliable in terms of performance metrics. Results based on evaluation also demonstrate that the hybrid algorithm is best one among the other classifiers which are used in this research and hybrid algorithm attained the highest accuracy (98.99) and the lowest false positive rate (1.01).