本文選題：認(rèn)知無線電網(wǎng)絡(luò) + 主用戶仿真攻擊��； 參考：《浙江大學(xué)》2017年碩士論文

【摘要】：認(rèn)知無線電網(wǎng)絡(luò)通過動態(tài)頻譜接入機制,可以有效利用空閑頻譜資源,緩解頻譜資源緊張問題。動態(tài)頻譜接入要求從用戶在不干擾主用戶網(wǎng)絡(luò)工作的前提下伺機接入頻譜,然而這種機制為認(rèn)知無線電網(wǎng)絡(luò)引入了新的安全問題。主用戶仿真攻擊(Primary User Emulation Attack,PUEA)是一類典型攻擊,惡意用戶通過模仿主用戶特征使從用戶錯誤判斷當(dāng)前頻譜已被使用,從而失去接入空閑頻譜的機會。PUEA問題干擾認(rèn)知無線電網(wǎng)絡(luò)的正常工作,尋找有效的PUEA防御策略是認(rèn)知無線電網(wǎng)絡(luò)安全的研究熱點�；诖�,論文研究基于信道特征的PUEA檢測技術(shù)。針對主用戶先驗信息已知的PUEA檢測問題,論文提出了一種基于信道多徑時延差的PUEA檢測方法。該方法采用多徑信道的傳播時延差作為檢驗統(tǒng)計量,并根據(jù)主用戶先驗信息與預(yù)設(shè)判決門限構(gòu)建二元假設(shè)檢驗,實現(xiàn)PUEA的檢測。同時,以虛警概率和漏警概率為性能指標(biāo),建模分析了所提出的PUEA檢測方法的性能。另外,利用通用軟件無線電外設(shè)搭建實驗平臺,在實際場景中驗證所提出的PUEA檢測方法的有效性。計算機仿真與實驗測試結(jié)果表明,基于信道多徑時延差的PUEA檢測方法能在滿足低虛警概率條件下,達到較高的檢測概率。針對主用戶先驗信息未知與環(huán)境參數(shù)變化的PUEA檢測問題,論文提出了一種基于增強學(xué)習(xí)的PUEA檢測方法。首先構(gòu)建基于單主用戶與多主用戶的認(rèn)知無線電網(wǎng)絡(luò)模型,并描述引入獎懲機制后頻譜內(nèi)各用戶的工作流程。然后根據(jù)不同系統(tǒng)模型,分析從用戶的獎懲收益情況,提出基于Q-Learning的PUEA檢測方法。該方法采用信道多徑時延差為狀態(tài)參數(shù),以判決門限為動作策略,優(yōu)化目標(biāo)為長時檢測收益(從用戶獲得的獎罰反饋),尋找不同環(huán)境下的最優(yōu)判決門限。仿真結(jié)果表明,基于Q-Learning的PUEA檢測方法能根據(jù)網(wǎng)絡(luò)環(huán)境參數(shù)變化實時調(diào)整判決門限,進而得到較好的檢測性能,有效提升從用戶的檢測收益。
[Abstract]:Cognitive radio network can utilize the free spectrum resource effectively and alleviate the problem of spectrum resource shortage through dynamic spectrum access mechanism. Dynamic spectrum access requires slave users to access the spectrum without interfering with the work of the primary user network. However, this mechanism introduces a new security problem for cognitive radio networks. Primary User Emulation attack is a typical attack in which malicious users misjudge that the current spectrum has been used by imitating the characteristics of the primary user. Therefore, the problem of missing the opportunity to access the idle spectrum interferes with the normal work of cognitive radio networks. Finding effective PUEA defense strategies is a hot topic in the research of cognitive radio network security. Based on this, this paper studies the PUEA detection technology based on channel features. In order to solve the problem of PUEA detection based on prior information of primary user, a PUEA detection method based on channel multipath delay difference is proposed in this paper. In this method, the propagation delay difference of multipath channel is used as the test statistic, and the binary hypothesis test is constructed according to the priori information of the primary user and the preset decision threshold to realize the detection of PUEA. At the same time, using false alarm probability and false alarm probability as performance index, the performance of the proposed PUEA detection method is modeled and analyzed. In addition, the experiment platform is built with the general software radio peripheral, and the effectiveness of the proposed PUEA detection method is verified in the actual scene. Computer simulation and experimental results show that the PUEA detection method based on channel multipath delay difference can achieve high detection probability under the condition of low false alarm probability. Aiming at the problem of PUEA detection with unknown priori information and change of environmental parameters of primary users, a PUEA detection method based on reinforcement learning is proposed in this paper. Firstly, a cognitive radio network model based on single master user and multiple master user is constructed, and the workflow of each user in the spectrum after the introduction of reward and punishment mechanism is described. Then, according to different system models, this paper analyzes the reward and punishment income of users, and puts forward a PUEA detection method based on Q-Learning. In this method, the channel multipath delay difference is used as the state parameter, the decision threshold is used as the action strategy, and the objective is to detect the income in a long time. (the reward and penalty feedback from the user is obtained to find the optimal decision threshold in different environments. The simulation results show that the PUEA detection method based on Q-Learning can adjust the decision threshold in real time according to the change of network environment parameters, and then obtain better detection performance, and effectively enhance the detection income from users.
【學(xué)位授予單位】：浙江大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TN925

【參考文獻】

相關(guān)期刊論文 前1條

1 張汝波,顧國昌,劉照德,王醒策;強化學(xué)習(xí)理論、算法及應(yīng)用[J];控制理論與應(yīng)用;2000年05期

相關(guān)碩士學(xué)位論文 前1條

1 吳偉;認(rèn)知無線電網(wǎng)絡(luò)中節(jié)能、安全的合作頻譜感知技術(shù)研究[D];浙江大學(xué);2011年

，

本文編號：1821482