Detecting colluders in pagerank: finding slow mixing states in a markov chain

  • Authors:
  • Benjamin Van Roy;Kahn Mason

  • Affiliations:
  • Stanford University;Stanford University

  • Venue:
  • Detecting colluders in pagerank: finding slow mixing states in a markov chain
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

The PageRank algorithm evaluates webpage reputations based on the hyperlinks that connect them. Webpages that collude to boost their reputations significantly distort the resulting rankings. We introduce a measure for assessing the degree to which a set of webpages boosts its reputation. There is no known efficient algorithm that is guaranteed to detect significantly boosted sets when they exist. However, we provide metrics that, under reasonable conditions, are guaranteed to detect a member of a significantly boosted set, if one exists, and address various implementation issues that arise in incorporating these metrics into PageRank.