Applying Long Short-Term Memory Recurrent Neural Network for Intrusion Detection

These days, web applications are used extensively. While organizations benefit from the new abilities they provide, the chance of being targeted is increased, which may cause massive system damage. It is thus important to detect web application attacks. Web intrusion detection systems (IDSs) are important for protecting systems from external users or internal attacks. There are however, many challenges that arise while developing a powerful IDS for unexpected and irregular attacks. Deep Learning approaches provide several methods, and they can detect known and unknown attacks. Long Short-Term Memory (LSTM) is a type of Recurrent Neural Network (RNN) and has the ability to remember values over arbitrary intervals. LSTM is a suitable method to classify and predict known and unknown intrusions. In this work, we propose a deep learning approach to construct an IDS. We apply LSTM RNNs and train the model using the CSIC 2010 HTTP dataset. An LSTM model using the Adam optimizer can construct an efficient IDS binary classifier with an accuracy rate of 0.9997.