BIG DATA SECURITY ENHANCEMENT BASED INTRUSION DETECTION SYSTEM USING K-MEAN CLUSTERING OF DECOMPOSITED FEATURES

Abstract

The protection of networked information systems is a critical problem impacting individuals, companies and governments. The number of attacks against networked networks has risen significantly and the methods used by the attacker are continuing to develop. Intrusion prevention is one method to avoid these threats from happening. A popular approach to creating an IDS system is by machine learning. The efficiency of the IDS is currently increased when discriminative and representative features are taken. AE and PCA are used to minimise dimensionality of features (PCA). The attribute extraction techniques employed are then used to construct an RF classification technique with K-Mean Cluster. This research effort will reduce the features of dataset "CICIDs" from 79 to 45, while retaining a high accuracy of 99.7% in Random Forest classifier with k-means clustering.