【摘要】：傳統(tǒng)的P2P（Peer to Peer）網(wǎng)絡(luò)被描述為可直接進(jìn)行資源與服務(wù)交換的網(wǎng)絡(luò)。P2P技術(shù)的核心思想是利用分布式存儲(chǔ)和計(jì)算能力來(lái)替代集中式處理方式。隨著移動(dòng)網(wǎng)絡(luò)與互聯(lián)網(wǎng)的融合，P2P網(wǎng)絡(luò)也必將進(jìn)入MP2P（Mobile P2P）網(wǎng)絡(luò)的新時(shí)代。MP2P網(wǎng)絡(luò)的主要功能是通過(guò)移動(dòng)設(shè)備間的直接數(shù)據(jù)交換來(lái)實(shí)現(xiàn)移動(dòng)服務(wù)的協(xié)同和資源的共享。由于MP2P能使資源受限的終端相互協(xié)同工作，并克服傳統(tǒng)P2P網(wǎng)中由于有限的寬帶和處理能力所帶來(lái)服務(wù)質(zhì)量下降的問(wèn)題，因此該網(wǎng)絡(luò)可以在較大范圍內(nèi)利用好有限的網(wǎng)絡(luò)資源。 移動(dòng)用戶間信任關(guān)系的確立是可信MP2P網(wǎng)首先要解決的關(guān)鍵性問(wèn)題。針對(duì)MP2P網(wǎng)與傳統(tǒng)P2P網(wǎng)終端設(shè)備在編址、通信方式以及標(biāo)識(shí)上的差異性以及網(wǎng)絡(luò)中可能存在的冒名、惡意詆毀、合謀以及“搭便車”等安全問(wèn)題，本文提出一種MP2P網(wǎng)下基于代理服務(wù)器的安全反饋信任模型（PSTM）：不同類別的代理服務(wù)器接入不同類型的終端，以屏蔽網(wǎng)絡(luò)層終端設(shè)備之間的差異性。同時(shí)，代理服務(wù)器之間對(duì)信息的相互備份能夠緩解服務(wù)器“單點(diǎn)失效”問(wèn)題。在資源安全選擇協(xié)議中對(duì)反饋方進(jìn)行身份、資格的認(rèn)證后進(jìn)行相似性篩選并加權(quán)處理。在多粒度的信任值計(jì)算中引入全局節(jié)點(diǎn)貢獻(xiàn)度和評(píng)價(jià)可信度，并將直接信任度分為面向節(jié)點(diǎn)與面向資源兩類，，來(lái)激勵(lì)移動(dòng)節(jié)點(diǎn)真實(shí)地反饋信息。實(shí)驗(yàn)表明：PSTM能夠減少詆毀以及合謀惡意行為，同時(shí)能夠抑制“搭便車”行為從而增加網(wǎng)絡(luò)善意節(jié)點(diǎn)的交易成功率。 在移動(dòng)節(jié)點(diǎn)無(wú)法獲取到代理服務(wù)器的全局評(píng)價(jià)可信度以及貢獻(xiàn)度的情況下，本文提出一種基于模糊理論的Q學(xué)習(xí)算法信任模型（FQTM），該模型研究的內(nèi)容是如何通過(guò)學(xué)習(xí)、適應(yīng)周圍陌生環(huán)境使得自己的利益最大化。FQTM將模糊理論應(yīng)用于Q-學(xué)習(xí)算法的狀態(tài)轉(zhuǎn)移概率的計(jì)算中，通過(guò)策略迭代算法不斷根據(jù)以往的交易對(duì)手調(diào)整自己的決策方案，更新Q值收益，最終Q值快速收斂至一個(gè)穩(wěn)定的狀態(tài)。實(shí)驗(yàn)表明：FQTM能夠使得移動(dòng)終端代理方的Q值收斂速度優(yōu)于貪心策略模型并且?guī)椭砣藢?shí)現(xiàn)利益最大化。
[Abstract]:The core idea of traditional P2P (Peer to Peer) network, which can exchange resources and services directly, is to make use of distributed storage and computing power to replace centralized processing. With the convergence of mobile network and Internet, P2P network will also enter the new era of MP2P (Mobile P2P) network. The main function of .MP2P network is to realize the cooperation of mobile services and the sharing of resources through the direct data exchange between mobile devices. Because MP2P can make resource constrained terminals work together, and overcome the problem of poor QoS caused by limited broadband and processing capacity in traditional P2P networks, the network can make good use of limited network resources in a large range. The establishment of trust relationship between mobile users is the first key problem to be solved in trusted MP2P networks. In view of the differences between MP2P network and traditional P2P network terminal equipment, the differences in communication mode and identification, and the possible security problems in the network, such as impersonation, malicious denigration, collusion and "hitchhiking", etc. In this paper, a security feedback trust model based on proxy server in MP2P network is proposed. Different types of proxy servers of (PSTM): are connected to different types of terminals to shield the differences between terminal devices in network layer. At the same time, the mutual backup of information between proxy servers can alleviate the problem of single point failure. In the resource security selection protocol, the feedback party is identified, the qualification is authenticated and the similarity is filtered and weighted. The global node contribution and evaluation credibility are introduced in the multi-granularity trust value calculation, and the direct trust is divided into node-oriented and resource-oriented to encourage mobile nodes to truly feedback information. Experiments show that: PSTM can reduce slander and malicious collusion, and restrain "free rider" behavior, thus increasing the transaction success rate of network bona fide nodes. When mobile nodes can not obtain the global evaluation reliability and contribution of proxy server, this paper presents a trust model of Q learning algorithm based on fuzzy theory, (FQTM),. The research content of this model is how to study by learning. The fuzzy theory is applied to the calculation of the state transition probability of the Q- learning algorithm, and the strategy iterative algorithm is used to adjust its decision scheme according to the previous counterparty, and the fuzzy theory is applied to the calculation of the state transition probability of the Q- learning algorithm. The Q value is updated and the Q value converges quickly to a stable state. The experiments show that: FQTM can make the Q convergence speed of mobile terminal agent faster than greedy policy model and help the agent to maximize the benefits.
【學(xué)位授予單位】：南京郵電大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：TP393.02;TN929.5

【參考文獻(xiàn)】

相關(guān)期刊論文 前7條

1 吳鵬;吳國(guó)新;方群;;一種基于概率統(tǒng)計(jì)方法的P2P系統(tǒng)信任評(píng)價(jià)模型[J];計(jì)算機(jī)研究與發(fā)展;2008年03期

2 高建清;王浩;于磊;方寶富;;一種模糊強(qiáng)化學(xué)習(xí)算法及其在RoboCup中的應(yīng)用[J];計(jì)算機(jī)工程與應(yīng)用;2006年06期

3 任艷;任平安;吳振強(qiáng);馬建峰;;移動(dòng)P2P網(wǎng)絡(luò)中的多粒度信任模型[J];計(jì)算機(jī)工程與應(yīng)用;2009年06期

4 陳學(xué)松;楊宜民;;強(qiáng)化學(xué)習(xí)研究綜述[J];計(jì)算機(jī)應(yīng)用研究;2010年08期

5 李小勇;桂小林;;大規(guī)模分布式環(huán)境下動(dòng)態(tài)信任模型研究[J];軟件學(xué)報(bào);2007年06期

6 歐中洪;宋美娜;戰(zhàn)曉蘇;宋俊德;;移動(dòng)對(duì)等網(wǎng)絡(luò)關(guān)鍵技術(shù)[J];軟件學(xué)報(bào);2008年02期

7 周文莉;吳曉非;;P2P技術(shù)綜述[J];計(jì)算機(jī)工程與設(shè)計(jì);2006年01期

本文編號(hào)：2259801