Research On Cross-technology Communication Security Technology Based On Machine Learning

At present,Io T technology has been widely used,and various heterogeneous Io T devices have been deployed on a large scale,playing an important role in smart cities,smart agriculture,smart homes,and smart transportation.Wireless communication technologies represented by Wi Fi,Zig Bee,and Bluetooth coexist and share the 2.4 GHz frequency band,causing wireless devices to compete for channels and interfere with each other.At present,cross-technology communication(CTC)enables heterogeneous wireless devices using different communication protocols to realize interconnection and interoperability without forwarding by a gateway,making data communication more convenient and greatly reducing the communication cost in heterogeneous networks.However,compared with the traditional homogeneous network mode,CTC technology also makes it easier to implement spoofing and congestion attacks in heterogeneous networks.With the help of longer communication distance and more sufficient energy supply,Wi Fi devices directly conduct spoofing and congestion attacks on Zig Bee devices,which brings great security risks to data communication under heterogeneous networks.This paper focuses on spoofing and congestion attacks based on CTC,as well as attack detection techniques,mainly using Wi Fi devices to spoof and congestion attack Zig Bee devices.For the CTC spoofing attack,we consider two situations to propose corresponding detection schemes:(1)Considering that during the experiment,it is difficult to fully cover the location of spoofing attacks.When there are no spoofing attack samples and the number of such samples is relatively small,there is an imbalance between the two types of samples(normal communication and spoofing attacks).The detection of CTC spoofing attacks can be achieved by using the outlier detection algorithm only by studying the single-type Received Signal Strength(RSS)data of Zig Bee devices.(2)When a large-scale spoofing attack breaks out in the network,two types of balanced samples can be obtained,and the deep learning-based binary detection model is used to detect CTC spoofing attacks.The main research contents are as follows:(1)Propose a CTC spoofing attack model and a CTC spoofing attack detection technology based on outlier detection.In order to ensure the communication security of heterogeneous Io T in CTC,in CTC,the outlier detection algorithm only needs to analyze the single-type received signal strength(RSS)data of normal communication to detect spoofing attack signals and realize timely identification,where RSS is important technical indicators that affect data identification in wireless communication.Set up three Wi Fi spoofing attack Zig Bee scenarios: the attack distance is less than 2m,the attack distance is greater than 2m and less than 3m,and the distance is greater than 3m and less than 5m.The RSS signal data were analyzed using outlier detection algorithms such as local outlier factor,isolated forest and single-class support vector machine(OSVM),respectively.By comparing the experimental results of the different anomaly detection algorithms above,it is found that OSVM has the best operating efficiency and detection performance in the three scenarios,thus determining the best OSVM-based CTC spoofing attack detection model.(2)Research on CTC spoofing attack detection technology based on deep feedforward neural network(DFN),which includes multiple stages,such as data preprocessing,data normalization,DFN-based classification method and parameter selection.In order to evaluate the performance of the DFN-based CTC spoofing attack detection model proposed in this paper in actual scenarios,the comprehensive evaluation indicators of decision tree,support vector machine,ensemble learning algorithms Ada Boost and XGBoost and the DFN model used in this paper are calculated respectively.The experimental results show that when the three attack distances are <2m,>2m and <3m,>3m and <5m,the DFN model has the highest AUC value in the comparison of the above five algorithms.Therefore,the DFN-based CTC spoofing attack detection model has good detection performance and supports spoofing detection for multiple attack scenarios.(3)Propose a CTC jamming attack model and its detection technology research.The traditional congestion attack model is usually: use the Zig Bee signal of one channel to congest other Zig Bee devices on the same channel.However,the CTC congestion attack model is: since the bandwidth occupied by the Wi Fi frame is 20 MHz,while the Zig Bee bandwidth is 2MHz,theoretically,using a single Wi Fi frame can congest the Zig Bee devices communicating in five different channels at the same time,and realize the anti-Zig Bee network.Communication maximum attack effect.In order to evaluate the performance of CTC congestion attack in real environment,this paper implements experiments based on USRP-N210 and MICAz hybrid platform.Finally,in order to detect the CTC congestion attack,a CTC attack detection experiment is carried out to compare the accuracy of the existing method based on signal strength consistency check and the method in this paper.The method improves the accuracy by an average of 90.8% over existing methods.