Application of Neural Network-Based Techniques to Network Intrusion Detection

Intrusion detection has been a challenge to safeguard computer networks. Several approaches have been proposed to make intrusion detection systems more effective in detecting intrusive attempts. In this work, we analyze the performance of four Neural Network-based techniques, namely Radial Basis Function Network (RBFN), Self-Organizing Map (SOM), Sequential Minimal Optimization (SMO), and LVQ 3 on intrusion data. Further, wrapper subset evaluator feature selection and standardization have also been considered for performance enhancement. The efficacy of the detection models has been evaluated using different parameters such as false alarm rate, negative predictive value, specificity, Matthew’s correlation coefficient, kappa statistic, and geometric mean.