Wi-FI Network Behavior Recognition Based On Physical Layer I/Q Signal

Cyberspace security is the focus of current security research,in which wireless network becomes the main target of network attack because of its openness and vulnerability to interference.Therefore,it is of great significance for wireless network intrusion detection.The traditional intrusion detection technology is aimed at the data information of the MAC layer and the network layer in the wireless network.The basic method is to create a fingerprint database of the normal behavior of known protocols,and compare the abnormal behavior to achieve intrusion detection.With the continuous change of intrusion detection requirements,there is an urgent need to identify network intrusion behaviors without relying on specific network protocols.In this paper,the method of network intrusion behavior recognition based on physical layer I/Q signal is proposed,and the method of machine learning is used to explore the network behavior recognition based on physical layer I/Q signal.This paper takes the wireless communication protocol IEEE802.11 g as the research object,collects wireless signals on the frequency band where the network is located,and classifies the IEEE802.11 g physical layer I/Q data with machine learning methods.The main research work of this paper includes:(1)Acquisition and processing of physical layer I/Q data: In the data acquisition stage,based on USRP and GNU radio software radio platform,design and implement the method for collecting and processing IEEE802.11 g physical layer I/Q data.(2)Feature extraction of data: In the feature extraction stage,design a dimensionality reduction model which combines the LSTM and the AE,and experiment to evaluate the parameter selection of the model.The feature extraction is realized.(3)Cluster analysis of data: In the cluster analysis stage,K-means model is implemented to cluster IEEE802.11 g physical layer signal data after feature extraction,and the clustering analysis is completed according to the clustering evaluation standard.This paper first proposes a framework for recognizing network behavior based on physical layer I/Q signals and machine learning.The experiments in real environment show that the proposed method can recognize network behavior without relying on the attacker's network protocol.Taking Wi Fi network as an example,the clustering accuracy reaches 76.23%,which verifies the effective classification of network behavior.