Neural KDE Based Behaviour Model for Detecting Intrusions in Network Environment

Abstract

Network intrusion is one of the growing concern throughout the globe about the information stealing and data exfiltration. In recent years this was coupled with the data exfiltration and infiltration through the internal threats. Various security encounters have been taken in order to reduce the intrusion and to prevent intrusion, since the stats reveals that every 4 seconds, at least one intrusion is detected in the detection engines. An external software mechanism is required in order to detect the network intrusions. Based on the above stated problem, here we proposed a new hybrid behaviour model based on Neural KDE and correlation method to detect intrusions.
The proposed work is splitted into two phases. Initial phase is setup with the Neural KDE as the learning phase and the basic network parameters are profiled for each hosts, here the neural KDE is generated based on the input and learned parameters of the network. Next phase is the detection phase, here the Neural KDE is computed for the identified parameters and the learned KDE feature value is correlated with the present KDE values and correlated values are calculated using cross correlation method. Experimental results show that the proposed model is robust in detecting the intrusions over the network.