130 likes | 253 Views
Trustworthy knowledge diffusion model based on risk discovery on peer-to-peer networks. Authors: Jason J. Jung Source: Expert Systems with Applications, vol. 36, pp. 7123-7128, 2009. Speaker: Shu-Fen Chiou( 邱淑芬 ). Outline. Introduction Proposed method Recommendation flow
E N D
Trustworthy knowledge diffusion model based on risk discovery on peer-to-peer networks Authors: Jason J. Jung Source: Expert Systems with Applications, vol. 36, pp. 7123-7128, 2009. Speaker: Shu-Fen Chiou(邱淑芬)
Outline • Introduction • Proposed method • Recommendation flow • Detecting malicious nodes • Experimental results • Conclusions • Comments
Introduction • Recommendation sharing • B like the computer information, when A owning a PCnews.doc, he wants to suggest B to download it. • High rating presents the item has more correlation from B. <x, rating information> A B x:PCnews.doc interest: computer information
Problem interest: computer information (invoker) True recommendation A B True recommendation False recommendation C True recommendation Malicious node True recommendation E D True recommendation
Requirements • Good recommendation propagation process • Detect malicious peer
Recommendation flow (interest x) (invoker) A B C E D
Two main drawbacks • Batch process • Assumed the majority of neighbors is fair and trustworthy
Conclusions • Attempting to model ‘‘word-of-mouth” recommendation propagation process in real world. • Filtering out the noisy (malicious) recommendation.
優缺點 • 優點:方法簡單。 • 缺點: • How to know the peer’s favorite items? • Only using recommendation to decide the malicious is not enough. • The performance in this paper is not good.
未來研究方向 • Not only using one item, using multiple items to decide whether exists malicious peer with a fix time. • Add the reputation matrix for every peer.