Internet of Things Intrusion Detection System Based on D-GRU
Yangchen Ge, Jian Li, Yuan Tian · 2022
Aiming at the shortcomings of slow training speed and insufficient accuracy of algorithms in IoT intrusion detection, an IoT intrusion detection system based on D-GRU neural network is proposed. The system uses window functions and D-GRU modules to extract the timing features in the network traffic and classify potential intrusions. The experiment was simulated on the intrusion detection dataset UNSW-NB15. In the binary-class and multi-class classification, the accuracy rates were 97.23% and 94.21%, respectively, and the parameters, training time, and prediction time were optimized in our scheme. The experimental results show that the IoT intrusion detection system based on the D-GRU algorithm maintains high accuracy while reducing the parameters, improving the performance of the system, and is more conducive to being deployed in resource-constrained edge nodes of the IoT.