Distributed Detection:Intelligent SSDF Attacks And Its Countermeasures

With the rapid development of wireless communication technology,wireless sensor network(WSN)has attracted more and more attention.Distributed detection is one of the important applications of WSN.The distributed detection system includes a target to be detected,several sensor nodes(SN)and a fusion center(FC).Each sensor node transmits the sensed measurement data to the FC,and then the fusion center processes the data and obtains related information of the target.In the distributed detection system,there appears Spectrum Sensing Data Falsification(SSDF)attack.Some sensor nodes are captured by an attacker to become malicious.The malicious nodes may send the falsified observation data to the FC,which influences the final decision of the FC and deteriorates the performance of the entire network.Therefore,it is significant to study the SSDF attack and its defense methods in distributed detection.Since most existing researches focus on eliminating the impact of malicious nodes,we analyze the impact of probabilistic SSDF attack on system performance from the perspective of attackers,and study the optimal attack strategy of malicious nodes from two practical goals.We mainly study the flip attack model and the offset attack model.Based on the KL divergence,the proportion of the smallest malicious nodes that "blind" the fusion center is obtained.The paper considers two attack goals.The first goal is to achieve the optimal attack strategy for the maximum damage performance.The second goal is to prevent distributed detection systems from working properly.The research results show that the first goal achieves optimal attack strategies to destroy the distributed detection system;the second goal achieves four optimal attack strategies,which not only break the balance of the system and make the system unable to work normally,but also make the malicious node unrecognized by the fusion center.In view of the existing malicious node identification methods,when the attack probability and attack strength of malicious nodes are small,the recognition probability of malicious nodes of these identification methods will be significantly reduced,and even malicious nodes cannot be identified.We propose a malicious node identification method based on adaptive threshold.The mismatch between the local decision result of the sensor node and the global decision of the system is used as the characteristic variable for sensor type judgment,and an adaptive threshold adjustment method based on attack probability and attack strength is designed.The research results show that this method can effectively identify malicious nodes not only when the attack probability and attack strength are high,but also achieve better recognition results when the attack probability and attack strength are low.Since most of the existing SSDF attack strategies remain unchanged in time,and there are few intelligent attack methods,we study the intelligent SSDF attacks and their defense methods.From the perspective of the attacker,the intelligent SSDF attack strategy is given.The malicious nodes aim to increase the false alarm probability and the missed detection probability,and consider the attack cost generated by the attack behavior,and obtain the optimal attack problem.By solving the optimal attack problem,the optimal attack probability is obtained.From the perspective of defenders,we propose a defense method that can identify intelligent SSDF attackers,which is improved based on the defense methods of the third chapter.The research results show that the intelligent SSDF attack strategy can achieve the maximum damage effect on the distributed detection system while ensuring that it is not recognized by the FC.The proposed defense method can also effectively identify intelligent SSDF attackers and mitigate the impact of attackers on the system performance.Using the SSDF attack strategies and its defense methods studied in this paper,from the perspective of the attacker,the malicious nodes can achieve effective attack;from the perspective of the defender,the fusion center can identify the malicious node and improve the system performance.