A Hybrid Model for Anomaly-Based Intrusion Detection System

Anomaly-based systems have become critical to the fields of information technology. Since last few years, evolution of anomaly-based intrusion detection system (IDS), improving detection accuracy, and training data preprocessing have been getting specifically important to the researchers of this field. In previous years, a lot have been discussed on the problems in using anomaly-based and hybrid IDSs. Anomaly-based approach is comparatively efficient from signature-based in novel attacks on computer network. However, in some cases, signature-based system is quick in identifying attacks from anomaly systems. In this work, authors have applied preprocessing in KDD 99 and have collected dataset using information gain. Authors have named collected dataset NUM15 as some of the features and redundant data are beside the point which decreases processing time and performance of IDS. After that, naive Bayes and Snort are used to classify the compression results and training the machine in parallel model. This hybrid model combines anomaly and signature detection that can accomplish detection of network anomaly. The results show that the proposed hybrid model can increase the accuracy and can detect novel intrusions.