Neural network & genetic algorithm based approach to network intrusion detection & comparative analysis of performance

In this paper backpropagation learning algorithm and genetic algorithm is applied for network intrusion detection and also to classify the detected attacks into proper types. During the training process of the backpropagation algorithm two possible set of features in the rule sets are used separately to determine proper rule set features for better performance. Then the performance of genetic algorithm is compared to the performance of both of the backpropagation approach. The process is tested on training dataset as well as test dataset to analyze the performance. It is found that in detecting the attack connections backpropagation algorithm shows better performance but in classifying the detected attacks into proper types the genetic algorithm approach is more successful.