Unsupervised Binary BAT algorithm based Network Intrusion Detection System using enhanced multiple classifiers

Network lntrusion Detection always faces challenges in creating a classifier that can handle the distribution of attack sequences in the KDDCup99 datasets. The attacks distribution in the training data is reasonable in the non-frequent categories compared to the entire population. As a result, categories and simultaneously, resulting in a poor detection rate. Selection of feature is a challenging task in wide range of data and choosing a customized subset of features always leads to better performance. To increase the detection rate of rare categories and improve overall performance, cascade classifiers are trained by using the partitioned training dataset and rare attack classes are separated by major categories. The “Binary Bat Algorithm” is used for feature selection technique. The proposed system has been implemented in the Hadoop framework by using MapReduce to improve execution time. The developed system is then explored with the classification algorithms such as the Naive Bayesian classifier and J48. The empirical result indicates that the proposed method achieves higher detection rates and high accuracy than existing methods.