Improving the anomaly detection by combining PSO search methods and J48 algorithm

• Main Authors: Kurniabudi, Kurniabudi, Abdul Harris, Abdul Harris, Mintaria, Albertus Edward, Darmawijoyo, Darmawijoyo, Stiawan, Deris, Idris, Mohd. Yazid, Budiarto, Rahmat
• Format: Conference or Workshop Item
• Published: 2020
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/92824/; http://dx.doi.org/10.23919/EECSI50503.2020.9251872
• Institution: Universiti Teknologi Malaysia

The feature selection techniques are used to find the most important and relevant features in a dataset. Therefore, in this study feature selection technique was used to improve the performance of Anomaly Detection. Many feature selection techniques have been developed and implemented on the NSL-KDD dataset. However, with the rapid growth of traffic on a network where more applications, devices, and protocols participate, the traffic data is complex and heterogeneous contribute to security issues. This makes the NSL-KDD dataset no longer reliable for it. The detection model must also be able to recognize the type of novel attack on complex network datasets. So, a robust analysis technique for a more complex and larger dataset is required, to overcome the increase of security issues in a big data network. This study proposes particle swarm optimization (PSO) Search methods as a feature selection method. As contribute to feature analysis knowledge, In the experiment a combination of particle swarm optimization (PSO) Search methods with other search methods are examined. To overcome the limitation NSL-KDD dataset, in the experiments the CICIDS2017 dataset used. To validate the selected features from the proposed technique J48 classification algorithm used in this study. The detection performance of the combination PSO Search method with J48 examined and compare with other feature selection and previous study. The proposed technique successfully finds the important features of the dataset, which improve detection performance with 99.89% accuracy. Compared with the previous study the proposed technique has better accuracy, TPR, and FPR.Anomaly Detection, CICIDS2017