Deep learning combined with de-noising data for network intrusion detection

Abstract

Anomaly-based Network Intrusion Detection Systems (NIDSs) are a common security defense for modern networks. The success of their operation depends upon vast quantities of training data. However, one major limitation is the inability of NIDS to be reliably trained using imbalanced datasets. Network observations are naturally imbalanced, yet without substantial data pre-processing, NIDS accuracy can be significantly reduced. With the diversity and dynamicity of modern network traffic, there are concerns that the current reliance upon un-natural balanced datasets cannot remain feasible in modern networks. This paper details our de-noising method, which when combined with deep learning techniques can address these concerns and offer accuracy improvements of between 1.5% and 4.5%. Promising results have been obtained from our model thus far, demonstrating improvements over existing approaches and the strong potential for use in modern NIDSs.