Research On Detection Methods Of Malicious Nodes In Cognitive Wireless Sensor Networks

The increasing use of wireless services and devices such as mobile communications,WIFI,and Internet of Things has made the radio spectrum resource more crowded.Cognitive radio(CR)provides opportunities to change the traditional static spectrum allocation framework.The main task of CR is spectrum sensing.cognitive wireless sensor network(CRN)consists of Multiple secondary nodes.Cooperative Spectrum Sensing(CSS)is a research hotspot currently focused on.CSS consists of multiple secondary nodes cooperating in sensing a specific frequency spectrum.When the primary user(PU)does not exist,the secondary node(SU)can opportunistically access the authorized spectrum hole to improving spectrum utilization.The biggest premise of CRN is not to interfere with PU.Due to the openness of wireless channels and cognitive radio,secondary users participating in CSS may be compromised by attackers and become malicious nodes.CRN faces both traditional network threats and CRN-specific network threats.Malicious nodes can cause degradation of network perception performance and low spectrum utilization.CRN-specific threats include spectrum sensing data falsification attacks and primary user simulation attacks.This paper analyzes the attacks of malicious nodes in CRN and the impact of the attacks on the network,and explains the attack behavior and defense strategies from the perspective of attackers and defenders,and proposes detection methods for malicious nodes corresponding to attacks.The attack of malicious nodes on CSS seriously affects the performance of spectrum sensing.Defense against malicious node attacks in CSS is a prerequisite for improving spectrum utilization and not causing interference to primary users.For the centralized network where binary hard decision is deployed,this chapter describes the behavior of malicious nodes in terms of attack probability parameter(?,?)and malicious node ratio ?,and proposes an algorithm to identify malicious nodes based on Sliding Window Reputation(SWR)algorithm.The SWR algorithm takes into account the frequency information of the hard decision 1 in the current sliding window and accumulates the reputation value in the manner of sliding window movement.This algorithm overcomes the shortcomings that the current user state estimation accuracy is low based on the current sensing slot,and the conventional algorithm is not suitable for networks with a majority of malicious nodes.In a network with a small number of malicious nodes,the SWR algorithm is compared with the algorithm in [28].Through simulation verification,the SWR algorithm is superior to the recognition algorithm in [28] on malicious node recognition performance.Using M-quantized data spectrum sensing is a new CSS method.This paper analyzes the attack model,optimal attack strategy,and blinding conditions under this framework,and analyzes the blind spot,Kullback Leibler(KL)divergence,and the proportion of malicious nodes Relationship.According to the unique attack strategy of malicious nodes in M-ary quantization,this chapter proposes a algorithm based on Symbol Expectation(SE)to identify malicious nodes.The algorithm identifies malicious nodes by the deviation between honest nodes,malicious nodes,and symbol expectations respectively.Through simulation verification,the SE algorithm performs better in performance metrics such as convergence rate,malicious node recognition rate,and threshold estimation complexity.