近年來網際網路迅速發展,線上社群與論壇的使用者日益增加,對檔案分享與資料傳送之需求也有增多趨勢。然而,P2P網路存在著一個主要的問題:大部分的節點對於在網路中的其他節點並不熟悉,在決定是否要從某節點下載檔案時,存在潛藏的風險。許多學者紛紛提出以信賴模型(trust-model)為基礎的解決方案,然而,先前的信賴模型皆為直接採取節點所推薦的分數來計算,因為每個人的評分習慣不同,如此將造成信賴值的誤差。 本論文提出一應用於P2P網路中:標準化推薦值之信賴模型,利用標準化(standardizing)推薦因子,此考量每個評分者過去的評分習慣,計算出他的評分區間範圍,再對應到一新的介於0到1之間的數值作為該次的推薦分數,另外,也考量其他因子,如:歷史因子、風險因子、獎勵因子,以上各因子分別配給相對應的權重,結算出一信賴值,以供節點在決定是否和某節點進行交易時做為一參考依據。 再者,本研究將與先前所提之MFTM模型進行分析比較,實驗模擬證明本研究所提之stdF Model 效能較佳。期望透過本論文所提之信賴模型,使每個節點的信賴值愈發準確,並能紓減目前在P2P環境中仍存在的問題,如節點上傳惡意檔案提供下載、過度放大本身評價等問題,進而提升P2P網路的服務品質。
With the blossom of peer-to-peer (P2P) applications, more and more people share resources through online community or forum. However, there is a major problem in the P2P environment - all nodes share resources using the same anonymous identity. There are potential risks when a user of a node decides to download files from the other nodes. Many solutions were based on a previous proposed trust-model which takes feedback and calculates trust values of nodes directly. However, it may have some calculation errors since behavior of evaluation differs from person to person. In this paper, we elaborate on a newly designed “a trust-model based on standardized feedback” (stdF model). It can calculate trust values of a node with standardized feedback factor, historical factor, risk factor, and incentive factor. After that, it will make a decision if a node is going to download files from another node or not. Moreover, performance of the proposed method is evaluated by simulation. Compared to MFTM model, results of simulations indicate that stdF model is much better.