Comparative Analysis for Detecting Malicious packets in LAN Network using Bagging and Boosting Techniques

Abstract

The security is the important one in sharing the data through the network. The intrusion detection system helps to classify the data as the normal or anomaly by using the machine learning algorithms. There is the large pool of machine learning algorithms which helps to find the anomaly. The algorithms have to find the connection that is normal or abnormal. The attacks are classified into four categories. Using the machine learning algorithm, the prediction is done to classify the packet as the normal or anomaly. The dataset used in this paper is communication of the LAN network which is happened in the very sensitive environments like military and Airforce. Various models are used to predict the anomaly but the bagging performs well in predicting the anomaly. There are various features associated in the dataset which helps to classify the network connection. The algorithms like adaboost was done with 10 iteration and the time for execution is 1.55 seconds and the accuracy are 94 %, bagging the time taken to execute the algorithm is 2.27 seconds, accuracy is 99%. Among these two algorithms the bagging performs well in predicting the abnormal data. Totally 42 attributes have been taken for the training and the testing purposes.