UAV Detection And Location Analysis Based On Wi-Fi Traffic

In recent years,unmanned aerial vehicles(UVAs)have been widely used in the civil field due to the development of technology and the reduction of manufacturing cost.UVAs was used in aerial photography,power patrol,rescue mission,agricultural monitoring,express logistics and other aspects.Although them can bring benefits to people,the threat caused is also gradually highlighted.A large number of them that are unauthorized may threaten public security and individuals’ privacy,such as the normal lifting of the airliner can be affected if the UAVs "black fly";photos of personal privacy,event meetings and outdoor exhibitions can be acquired if the UAVs are abused.In order to minimize the threat caused by unauthorized UAVs,there is an urgent need to develop technical means used for control them.At present,the detection and tracking of UAVs mainly have the following difficulties: 1.Due to the wide variety of small drones and fast flight speeds,so it is difficult for us to detect the presence of unauthorized them in a designated area and in a timely and effective manner;2.In the complex environment,such as metropolis,large positioning equipment(such as radar)cannot be deployed.As a result that it is difficult to grasp location of unauthorized UAVs timely,and it is impossible to perform directional interference on them.Aiming at the above problems,this paper conducts research on the identification and positioning of UAVs based on Wi-Fi traffic.The following work is done in this paper: 1.The first trouble: timely and effective identification of unauthorized UAVs.This paper designs a method based on wireless device configuration information recognition and a method based on the statistical characteristics of Wi-Fi packets.In order to shorten the time of unauthorized UAV identification,this paper captures the Wi-Fi data packets between UAV and its controller,extracts tuple information,and comprehensively uses the configuration information of UAV manufacturer to the AP module of UAV,such as MAC address,BSSID,port and other information to determine whether the captured data packets belong to UAV.The results are divided into yes,no and still in suspected,and for the suspected data packets and encrypted Wi-Fi data,further identification is carried out,using the identification method based on the statistical characteristics of Wi Fi data packets.In the identification method based on the statistical characteristics of Wi-Fi data packets,the accuracy of identification increases with the increase of number of captured data packets.In order to reduce the prediction time of the data packet,a balance is reached between the time cost of prediction and the accuracy of recognition.A reweighted machine learning model needed to be designed which considers the time of sample collection,the calculation time and accuracy of features,and performs packet-by-packet analysis of captured packets,to reduce the time of UAVs identification.2.The second trouble: it is difficult to grasp the location of unauthorized UAVs in complex environment such as metropolis timely.In this paper,we use the relationship between received signal strength indication(RSSI)and distance model and fingerprint matching algorithm based on weighted nearest neighbor.First,RSSI is a very unstable measurement index,which is affected by many factors,so outlier detection algorithm and filtering are used to preprocess the collected RSSI to overcome the impact of the environment on the signal RSSI.And then relatively stable RSSI value can be obtained and a stable offline position fingerprint database can be established.Secondly,in the online positioning phase,the positioning method is adopted by using the anchor point as the center of the ball to intersect at one point.However,in the case of weak adaptability,the localization range is limited.Therefore,the positioning method is extended.For the case where it cannot intersect at one point,the centroid-based positioning method is used.For some anchor points that do not receive the RSSI value of the signal from the position to be measured,a position-based fingerprint matching method is used.And through an experiment,achieve the expected effect.