Intrusion Detection System Using Enhanced AdaBoost Algorithm

Intrusion detection system is the process of detecting computers or networks for unauthorized activities or file modification. It can monitor network traffic and detecting the network attacks such as Denial of service, Probe, R2L, U2R Attacks. More specifically, Intrusion detection. system tools to detect specific computer attacks and unauthorized use of system, and to alert the proper users usually Network Administrators about detection of attacks in system. These techniques help in them determining what qualifies as an intrusion (attacks) versus normal. The Intrusion detection can be done through a Cascading classifier approach which is used to increase the detection rate of the rare network Classification Algorithm. At First, the Training set network data are sent to both algorithm for learning purpose. Bayesian network detects rare attacks in a better way. So those data records which are classified as a "NORMAL" by J48 Model network attack categories like R2L, U2R Attacks. Here J48 Classification algorithm was cascaded with Bayesian are sent to Bayesian Network Model and Classified. This result in a better detection rate of rare network attacks. This paper explains the splitting of KDD training sets into two sets and we use Particle Swarm Optimization for selectingimportant features from two training sets for learning and then Enhanced Adaboost (Ensemble) classifier was cascaded with Bayesian Network Classifier. This algorithm repeatedly calls weak learner (classifier) to increase classification accuracy rate. This result in increasing the detection rate of both rare and non-rare network attack categories.