Intrusion Detection System using MLP and Chaotic Neural Networks

Network intrusion detection system is a process that attempts to determine the abnormalities in system or network activities by observing traffic of the network to identify malicious activity correctly. Many challenges arise while developing a flexible and efficient NIDS for unexpected and impulsive attacks. Statistical, machine/deep learning based techniques are used to design an intrusion detection system. However, most of the work is focussed on improving the accuracy of the model ignoring the false alarm rates. In this work, we use a hybrid model of Multilayer perception and chaotic neural networks to improve the accuracy, precision as well as the false alarm rate. MLP is used to detect the attacks and chaos neural networks are used to reduce the false alarm rates. We use KDD Cup'99 a benchmark dataset for experimental analysis. The results reveal that Hybrid approach performs better than the MLP approach in terms of reducing false alarm rates with similar performance in accuracy, precision and recall.