Network Intrusion Detection System Using Machine Learning Algorithms

Abstract

The fast progress in the sectors of web and communication led in an enormous growth in network size and content. As a consequence, several new attacks are produced and network security problems to discover invasions properly. In addition, invaders cannot be overlooked in order to start several assaults inside the infrastructure. One instrument that safeguards the system from probable infiltration through the inspection of network traffic to assure its secrecy, completeness and accessibility is an infiltration identification system. Current systems of passive identification capture only recognized malicious attacks and also regular modifications in datasets based on signatures. To decrease the task, network infiltration identification systems (IIS) are presented that are able to evaluate network contents using machine learning (ML) approaches for evaluating and classifying hazardous items. Different ML algorithms are employed, such as k-means, K-means with Principal Component Analysis (PCA), extreme learning approaches, Naive Bayes and Hoeffding Tree algorithm, Random decision Forest algorithm. In addition, Accuracy Updated Ensemble method, accuracy weighted ensemble algorithm, supports the development of a Network Infiltration Identification System by a vector machine. Some of these methods are now tested to and evaluated according to the NSLKDD data set.