Deep Learning with Encoders for Intrusion Detection Systems (IDS)

Abstract

An Intrusion Detection System (IDS) is often used to keep safe and efficiently utilise communication network. And generally this is achieved by monitoring the network regularly for possible intrusions. Nowadays in networks, zero-day attacks are common type of intrusions. So it became necessary to use IDS, which is capable of detecting zero-day attacks. Machine Learning (ML) algorithms are suitable in designing such IDS. Deep Learning (DL), being part of ML, have promising approaches that can be used in designing IDS, which is capable of detecting intrusions of new and old types. This paper proposes DL based IDS which create model of normal network traffic. This model classifies attacks based on the various features present in the dataset. The proposed model uses a Convolution Neural Network, bi-directional Long Short Term Memory (LSTM), and a stack of encoders to handle spatial and temporal features more effectively. The model achieves a low False Positive Rate and high Detection Rate compared to existing models.