A Distributed P2P Link Analysis Based Ranking System

Link Based approaches are among the most popular ranking approaches employed by search engines. They make use of the inherent linkage based structure of World Wide Web documents assigning each document an importance score. This importance score is based on the incoming links for a document; a document which is pointed to by many high quality documents should have a higher importance score. Googles’ highly popular search technology [1] exemplifies the success of the link based ranking algorithms in identifying important pages. However, such link analysis based algorithms suffer from some drawbacks. Googles’ PageRank algorithm has an update time extending into months which is not feasible for frequent updating of the system. Secondly, the algorithm is susceptible to manipulation by malicious Web Spammers, who manipulate the link based analysis to favor their fake websites. The problem commonly termed as Web Spam can seriously hurts the performance of PageRank algorithm, leading the algorithm into providing unjustified high PageRank to spam web pages. In [2], the authors propose the SourceRank approach for enhancing PageRank through source-based link analysis, which can potentially help combat the problem of web spam. In this project, we propose to implement the SourceRank technique on top a P2P crawler Apoidea [3]. We plan to perform experimental analysis using the SourceRank technique on the www.gatech.edu domain and analyse the results to verify the claims made in [2]. In addition, we analyze the results obtained by the above experimentation to enable us to answer an important question raised in [2] – How can we identify a collection of pages which constitutes a single source?