相較於主從式架構,從點對點網路相關的應用系統的熱門程度以及多樣性可以觀察得到,其獨特的網路架構為資訊分享帶來的強大優勢。然而在和點對點網路相關的應用之中,由於欠缺鼓勵合作的誘因(incentive),讓大多數的使用者不願意分享其資源,於是造成其習於只使用卻不貢獻資源的這樣搭便車(free riding)的不良行為。當此搭便車的行為盛行於點對點網路中,將會和傳統主從式架構一樣的產生效能和安全上的問題。一種解決此種困境,亦即產生使用者分享、合作的誘因的方法,是透過透使用者不同的貢獻程度來差異化其服務水準,本研究即是為點對點網路提出一種結合虛擬社群(virtual community)以及模糊邏輯(fuzzy logic)等概念的信譽系統(reputation system),稱之為FuzRep。利用虛擬社群的成員之間密切互動所產生的內聚現象,來達到社群成員之間彼此監督、彼此鼓勵的信譽管理機制。本研究強調此系統的信譽管理機制為非集中式的,同時也不限制底層的點對點網路必需以何種結構為基礎。同時,FuzRep亦將使用者的服務差異化視為一種決策過程,因此採用了一種模糊邏輯的機制來幫助此決策過程。本研究利用系統模擬的實驗來探測此信譽系統的運作情形,最後的實驗結果也顯示了此種設計的可行性。
Peer-to-Peer (P2P) networks have shown their great strength as an information sharing architecture what we witness from the popularity and the variety of P2P-related applications. Lack of incentives, however, makes most peers unwilling to cooperate and lead to free-riding behavior. The prevalence of free riders will, eventually, bring P2P networks performance and security issues as what traditional client-server architecture encountered. One way to encourage cooperation is through service differentiation based on each peer’s contributions. This paper presents FuzRep, a reputation system that incorporates virtual community and fuzzy logic techniques for P2P networks. Without using information hubs or DHT-like pre-sorting structure, we utilize the clustering effect of virtual communities for reputation management. Furthermore, we treat service discrimination for resource request as a decision-making process so that a fuzzy-based mechanism is adopted in FuzRep to facilitate that process. Finally, the experimental results of this reputation system positively indicate the feasibility of our design.