Intrusion Detection of IoT Traffic Payload Based on Parallel Neural Networks

A malicious traffic intrusion detection method based on packet payload is proposed to address the current flow based intrusion detection models that overly rely on statistical information from packet headers and ignore the characteristics of packet payloads. Firstly, by using the packet payload labeling algorithm, the labeled payload data is obtained, and then the data is cleaned and standardized; Input it into the constructed composite neural network, use parallel neural networks to extract spatial features of payloads, and finally input the extracted features into the fully connected layer for classification. To demonstrate the feasibility of this research method, experimental results on the UNSW-NB15 dataset of the Internet of Things showed that the method is practical and feasible, achieving a higher F1 score compared to the comparative method.