Research On Passive Intrusion Detection Method Based On Internet Of Things

Passive intrusion detection technology can overcome the requirements of the line of sight compared with the traditional infrared,image and video detection technology,and does not need the testee to carry any special hardware equipment.It has abstracted great attention in military security,animal research,personnel monitoring,energy saving,smart home and other fields.With the rapid development of Internet of Things(IoTs),a variety of equipment in IoTs have entered many aspects of life.It is possible to help traditional intrusion detection system reduce the dependence of special hardware and improve the applicability and popularity by using the existing passive intrusion detection system.The traditional passive intrusion detection method is based on the statistical characteristics of the wireless signal,requires a lot of data and complex training.Because the effect of environmental changes,the detection results are not ideal.How to design a passive intrusion detection system with high accuracy and low false alarm rate based on IoTs is still a challenging problem.In this paper,based on the existing passive intrusion detection system,the intrusion behavior detection is realized by using the general network parameters: Received Signal Strength Indication(RSSI)in the mainstream IoTs system.Based on the characteristics of RSSI data,the characteristics of IoTs,as well as the characteristics of the working environment of the system,this paper have designed specific solutions to achieve passive intrusion detection.The passive intrusion detection system in this paper mainly includes RSSI acquisition stage,learning stage and dynamic optimization stage.The main steps are: packet acquisition,RSSI data extraction,feature extraction,feature clustering,discriminant analysis,dynamic optimization.Firstly,the RSSI data are collected in the environment and analyzed by Horizontal Hierarchy Slicing(HHS)algorithm based on the mathematical morphology and the corresponding morphological characteristic curves can be generated.This method can solve the problem that the statistical characteristics of signals are not stable.Then,based on the Partitioning Around Medoids(PAM)algorithm to do the clustering analysis of the feature data.PAM algorithm does not need to be trained in the model.It is easy to deployed and it candeal with the noise,isolated points and other anomalies in the data set.Then the discriminant analysis module use the improved Agglomerative Hierarchical Clustering(AHC)algorithm to classify the data acquired in real time.The discriminant function can discriminate the real-time data whose class is unknown based on the existing class and output the result to the users.Finally,in order to improve the environmental adaptability of the detection system,an optional manual optimization step is set up to help the passive intrusion detection system update the cluster category and improve the performance of the detection system.In this paper,the experimental details of the study process are introduced in detail and the experimental results are analyzed detailedly.In order to verify the versatility and effectiveness of the system,RSSI data are collected based on Bluetooth,Zigbee and Wi-Fi communication systems.Each system has been deployed in three different physical environments.Through the concrete experiment simulation,it is verified that the passive intrusion detection method proposed in this paper can effectively detect the intrusion in the environment with good environmental adaptability.