Network Intrusion Detection using Neural Network Based Classifiers

摘要

Rapid expansion of computer networks throughout the world has made data security a major concern. In the recent past, there have been incidences of cyberattacks which have put data at risk. Therefore, developing effective techniques to secure valuable data from such attacks is the need of the hour. Several intrusion detection techniques have been developed to deal with network attacks and raise alerts in a timely manner in order to mitigate the impact of such attacks. Among others, ANN methods can provide multilevel, multivariable security system to meet organizational needs. In this work, we have applied four prominent neural network based classification techniques, viz., SelfOrganizing Map, Projective Adaptive Resonance Theory, Radial Basis Function Network, and Sequential Minimal Optimization to predict possible intrusive behavior of network users. The performance of these techniques have been evaluated in terms of accuracy, precision, recall / detection rate, FMeasure, and false alarm rate on the standard NSLKDD intrusion dataset.