Network Intrusion Detection by Finding Correlation between Multiple Features using K-Means Algorithm & Multivariate Correlation Analysis

Network attackers are turning towards interconnected systems in networks such as web servers, database servers and cloud computing servers, thus the availability and reliability of network services are being intimidated by the growing number of Network Intrusion attacks. Effective mechanisms for Intrusion detection are needed. Different systems were proposed to safeguard these systems from intrusion attack using machine learning, statistical analysis, data mining, etc. The proposed system is an improvement over earlier systems as k- means clustering is applied to train samples so that it can categorize the samples into different clusters. After which it applies statistical methods to find the correlation between features which are clustered to gain statistical numbers in form of standard deviation, mean and covariance matrix. Applying the multivariate correlation analysis on each cluster will help us to get profile parameters according to their cluster and thus allowing us to know sharp boundary of characterizing a sample packet. This in turn will help us to reduce false positive rate.