| Since the emergence of wireless communication technology,it has been rapidly developed and widely used in our lives,such as radio and television,aerospace,mobile communications and other fields.The limited spectrum resources are increasingly strained.However,in the actual use process,many idle spectrum has not been fully utilized.In response to this problem,some scholars proposed Cognitive Radio(CR)technology,it allows secondary users to use idle spectrum in authorized channels to meet their communication needs without disturbing the normal communication of the primary user.Spectrum sensing is a key step to discover free spectrum in cognitive radio technology.Cooperative Spectrum Sensing(CSS)can effectively improve spectrum detection capability,but cooperative spectrum sensing introduces malicious users,among which Spectrum Sensing Data Falsification Attack(SSDF)attacks are the most widespread.Malicious users affect the overall judgment of the system and reduce detection performance by forging perception data and uploading false perception results.This thesis will focus on the SSDF problem in centralized networks,and the main research work is as follows:1.In a single channel environment,this thesis proposes an Adaptive Trust Threshold defense algorithm based on Reinforcement learning(ATTR)to detect malicious users against both ordinary SSDF attacks and intelligent SSDF attacks.Due to the fact that Intelligent Malicious User(IMU)can dynamically upload real perception results to play the role of honest users when launching intelligent SSDF attacks,traditional trust detection mechanisms cannot accurately identify them.This thesis introduces a combination of reinforcement learning and trust mechanisms,and sets reasonable trust thresholds for each Secondary User(SU)based on their perception data.By using adaptive trust thr esholds to increase the attack cost of malicious users,malicious users can be filtered out and a collection of honest and trustworthy users can be constructed.The simulation results show that the defense algorithm proposed in this thesis can improve the system’s recognition rate of trusted users,effectively defend against ordinary SSDF and intelligent SSDF attacks,thereby improving the accuracy of the final decision results and improving the detection performance.Under different SSDF attack intensities,a detection probability of 90%can be achieved quickly.2.In a multi-channel environment,this thesis proposes a multichannel based SSDF defense and channel detection sequence algorithm to address the issue of channel detection sequence in multi-channel SSDF attacks and cooperative spectrum sensing.SSDF malicious users in multichannel environments may exhibit different honesty characteristics towards different channels to evade detection mechanisms.In order to defend against this issue,this thesis uses an improved ATTR strategy to establish corresponding trusted user sets for different channels,to defend against multi-channel SSDF attacks and improve the accuracy of channel state detection.At the same time,in a multi-channel environment,due to the limited number of channels that secondary users can detect in each time slot,the issue of channel detection order needs to be considered.This thesis models it as a Multi-Armed Bandits(MAB)problem,and predicts the idle probability of the channel through Thompson sampling to optimize the channel detection order.The simulation results show that the proposed algorithm can filter out honest and trustworthy user groups for each channel,improve the defense ability against multi-channel SSDF attacks,and improve the prediction accuracy of channel idle probability,thereby improving channel detection efficiency. |
