Host-based Intrusion Detection System (HIDS)

This paper presents the data analysis and feature extraction of KDD dataset of 1999. This is used to detect signature based and anomaly attacks on a system. The process is supported by data extraction as well as data cleaning of the above mentioned data set. The dataset consists of 42 parameters and 58 services. These parameters are further filtered to extract useful attributes. Every attack in the dataset is labeled either with “normal” or into four different attack types i.e. denial-of-service, network probe, remote-to-local or user-to-root. Using different machine learning algorithms, the work tries to compare the individual accuracy, True Positive and False positive rate of every algorithm with every other algorithm. The work focuses its attention to increase security through detection of static as well as dynamic attack.