FEATURE SELECTION BASED ON MAXIMAL INFORMATION COEFFICIENT FOR INTRUSION DETECTION SYSTEM

• Main Author: UTAMA PUTRA (NIM: 23215061), AGUNG
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/25131
• Institution: Institut Teknologi Bandung

With the development of technology and the speed of network-based applications, the threat of cybercrime also increases. One of the methods used to prevent the occurrence of such attacks is to use an intrusion detection system. IDS collect and analyze different areas of the computer and network to identify attack attempts that could compromise the confidentiality, integrity and availability of systems and networks. In the building of IDS, the main problem of concern is the data dimensions that needs to be analyzed. Feature selection is an important step in building an IDS to select a subset of features that are small enough but still informative enough to characterize traffic, thereby reducing data dimensions and shortening development time. In this thesis, a feature selection method based on Maximal Information Coefficient for IDS is proposed. The proposed feature selection method will also be compared to the commonly used mutual information-based feature selection method. The proposed feature selection methods are tested using the UNSW-NB15 data set. The analysis shows that the selection of MIC-based features provides performance equally well with the selection of feature-based mutual information for intrusion detection systems, but with fewer number of features and therefore succeeding in reducing more development time of the IDS.