Intrusion detection using neuarl networks trained by differential evaluation algorithm

Nowadays Information security is an important issue in Information Technology world. The computer viruses, worms, hackers, crackers, electronic eavesdropping and electronic fraud, intrusions are some of the problems that Computer Security experts are facing. The Intrusion Detection System is a common and widely used approach in a well formed network security policy. Information systems must be monitored and audited for potential attacks; but the challenge in this process is analyzing heavy loads of event logs and network traffic. Also to be able to recognize new kinds of threads that tack place in network every day in a timely and efficient manner. In this paper we considered Differential Evolution algorithm for training neural network for the intrusion detection system. We used KDD dataset for our experiments that is derived from the standard KDD CUP" Intrusion Dataset. We also provided the comparative results of the differential evolution with the state of the art classification algorithm like RBF, Probabilistic Neural network (PNN) and Multilayer Perceptron (MLP) neural network. We reduced the dimension/features of the KDD datasets using PCA. The results of our study showed higher accuracy in intrusion detection.