Anomaly Detection in Network Traffic using Deep Learning

Abstract

Anomalous behavior of network traffic indicates an underlying intrusion or malicious intent at play. Various techniques are available to detect anomalies like signature-based techniques, statistical methods and rule-based techniques are a popular choice. All these above stated techniques suffer from issues of detecting novel threats. As the amount of data transmitted over the internet keeps growing, there is a growing need for intelligent internet safety solutions. In this paper the work will focused on deep learning solutions which are capable of learning complex patterns in the network traffic. The use of autoencoders will be made which are able to capture both continuous and non- continuous patterns in data and then utilize reconstruction error to look for irregularities.