Research On Spectrum Sensing Of Cognitive Radio Against SSDF Attack

Nowadays,the rapid development of wireless communication makes the shortage of spectrum resources more serious.However,the fixed spectrum allocation mode causes most of the spectrum to be unused,and the utilization of spectrum resources is low.Traditional methods to improve spectrum utilization(such as multi-carrier transmission,multi-antenna technology,etc.)have not fundamentally solved the problem of spectrum resource shortage.Cognitive radio technology makes use of the free spectrum opportunity to communicate effectively and improves the spectrum efficiency by spectrum sensing without causing harmful interference to the existing authorized users.Spectrum sensing is the most important part of cognitive radio technology.Centralized cooperative spectrum sensing is widely used because of its fast sensing speed.If there are malicious attacks(such as SSDF attacks)in sensing,the system performance will be seriously affected.Therefore,this paper focuses on the research of SSDF attacks in cognitive wireless networks to improve the detection performance of the system.Firstly,when a small number of malicious users take static attacks,the detection performance of the traditional abnormal data processing algorithm decreases when malicious users exist.In this paper,we propose an algorithm against SSDF attack based on K-Means and improved PSO.Firstly,in the aspect of malicious user identification,the traditional anomaly data detection algorithm is inaccurate in the presence of malicious data.In this paper,OGK criterion is used to initialize the clustering center to get more accurate classification results.Secondly,in the aspect of malicious data processing,this paper proposes a method of combining particle swarm optimization algorithm with SNR to allocate weights reasonably to cognitive users for final fusion decision.The simulation results show that the proposed algorithm has good detection performance and effectively resist SSDF attacks in the case of malicious users launching attacks.Secondly,when most malicious users take static attacks,the traditional evidence-based algorithm can not accurately evaluate the differences between cognitive user data and can not accurately identify the malicious user when the reported sesning data is approximate.In this paper,we propose an algorithm against SSDF attack based on evidence theory and fuzzyentropy.Firstly,the conflict coefficient is redefined according to the distance of evidence and the classical conflict coefficient,and the conflict weight is given to the user's sensing data.The fuzzy weight of the sensing data is obtained by the fuzzy entropy.The trust weight is calculated according to the user's current sensing results.The basic probability assignment function is modified by the above three parts of weights,and the fusion center obtains the final sensing result according to the fusion formula of evidence theory.Simulation results show that the proposed algorithm has better detection performance than other algorithms,and solves the problem of poor performance of most malicious users launching SSDF attacks.Finally,when most malicious users take dynamic attacks,the traditional reputation mechanism does not have a reasonable evaluation of the data sensed by the cognitive users.When the malicious users launch the SSDF attack,the detection performance of the system drops.This paper proposes a dynamic trust mechanism based on Beta reputation system.Firstly,the credit value of cognitive user is initialized.Secondly,assign the reputation value to the user using the Beta reputation system according to whether the user correctly sensing.Then,record the user's historical sensing results continuously,obtain the users' continuous honesty and false sensing times and get the rising factor and the falling factor through certain processing.Update the user's reputation value and get the final fusion decision after gaining weight.Simulation results show that the trust algorithm has lower false alarm probability and missed detection probability when the proportion of malicious cognitive users is high,and still has good spectrum sensing performance.