| Wireless Sensor Networks(WSN)consist of a large number of low-cost,low-power,mobile miniature sensor nodes,and have a wide range of promising applications in military and civilian fields such as environmental monitoring,target detection,and battlefield surveillance.WSN-based distributed detection has received great attention as an important application in the research of WSN.The harsh deployment environment of WSNs and the openness of wireless channels make them highly susceptible to security threats such as malicious attacks.Therefore,it is increasingly important to resist malicious attacks,design secure and effective distributed detection methods,avoid information leakage,and ensure WSN detection performance.In view of this,this thesis addresses the eavesdropping attack and Data Falsification Attack(DFA)problem in WSN-based distributed detection,and carries out distributed detection performance analysis,secure distributed detection algorithm research and simulation experimental verification,etc.The main research results are summarized as follows:(1)To address the problem of eavesdropping attacks on WSNs,first,considering the distributed detection network model based on WSNs and global eavesdropping attacks,we theoretically analyzed that when the local operating point of a sensor is close to the receiver operating characteristic(ROC)endpoint,the fusion center(FC)and the quality of the sensor decisions received by the eavesdropper(Eve).Second,in the NeymanPearson(N-P)framework,when the local sensor and FC are noisy communication channels,a compromise analysis of detection performance and security performance is proposed using the Kullbaci-Leibler Divergence(KLD)ratio of FC and Eve,and simulation results show that the method achieves the detection performance under the upper detection performance constraint obtained for a given Eve,i.e.,the safety performance constraint.(2)For the global eavesdropping attack problem faced by WSNs,when the local sensor and FC are noiseless communication channels,the theoretical analysis shows that:asymptotically perfect security performance and asymptotically perfect detection performance can be obtained by increasing the number of sensors;the detection performance limit obtained by FC is theoretically analyzed under the Bayesian framework with the global error probability in FC and Eve as the performance index;the derivation The approximate asymptotic error indices of FC and Eve under the security constraint are derived,and the simulation verifies the achievability of asymptotic perfect detection under the security performance constraint.(3)To address the DFA problem faced in WSN-based distributed detection,this thesis introduces the idea of social learning into distributed detection and proposes a social learning-based secure distributed detection method.The method uses tandem ranking decision making to make judgments sequentially in random order based on the local observations of sensor nodes and the received social information transmitted by other nodes,and derives data fusion rules and judgment criteria.The mathematical characterization of the information cascade under DFA is given,and the simulation results show that the proposed method can effectively resist DFA attacks and improve the robustness of the WSN network under DFA attacks. |
