A deep learning approach based on sparse autoencoder with long short-term memory for network intrusion detection

In recent years, a deep neural network has been solving a variety of complex problems of science and engineering fields ranging from healthcare to transportation. Among them, one of the most crucial issues is to protect a network against cyber threats. In this article, we present a two-stage IDS framework based on a single-layer Sparse Autoencoder (SAE) and Long Short-Term Memory (LSTM), to design an effective network intrusion detection. Initially, the single-layer SAE learns new feature representations of the data through the nonlinear mapping, following that, the new feature representations are fed into the LSTM model to classify network traffic whether it is being normal or attack. The proposed framework was evaluated on the benchmark NSL-KDD dataset, where the mean accuracy of the proposed method was achieved 84.8%. The experimental results show that the two-stage IDS framework achieved better classification accuracy than the existing state-of-the-art methods. © 2020 The Institute of Electrical Engineers of Japan.