Bidirectional LSTM based approach for network intrusion detection

Abstract

In today’s digital age the number of people that have access to the internet services have grown in leaps and bounds. It has been directly associated with the increasing security threats. Traditionally it has been difficult to handle these security threats, but with the recent advancements in recurrent neural network architectures have paved a path for effective threat identification. Intrusion detection system has been a widely researched area in security analysis and evaluation. In this paper, we are leveraging the Bidirectional LSTM (Long Short Term Memory) networks for Intrusion detection model. Bidirectional LSTM uses all the available information in the network and also provides context to the network than the normal LSTM. NSL-KDD dataset is used to validate the proposed model. To evaluate our proposed model, we have compared our model against vanillaRNN, normal LSTM and GRU networks which are most popular models for network intrusion detection. They are compared in terms of accuracy, precision, recall and F1-measure.