Deep Learning for IoT Intrusion Detection based on LSTMs-AE

Abstract

With the advent of 5G era, The Internet of Things (IoT) is obtaining considerable attention in all walks of life nowadays. However, due to the hardware problems of devices, there may exists some security problems in IOT. While existing intrusion detection methods rarely consider the time series feature of the data. In this paper, we propose an anomaly monitoring model for Autoencoder based on Long-Short Term Memory (LSTMs-AE), in which LTSM is exploited to capture time-series features and the intrusion detection is performed by the feature learning ability of Autoencoder. Thorough experiments demonstrate that our model has better intrusion detection performance than ordinary Autoencoder, as in most of the dataset the accuracy rate of proposed scheme exceeds 0.95.