|
In some P2P file sharing systems like BitTorrent, portal web pages with different web features are often adopted to guide resource selection. We believe the features on these pages will directly affect the users’ decisions of selecting resources, which further influence the following process of content delivery. We conducted an in-depth measurement of “5QZone”, the largest BitTorrent website in CERNET. Our crawler has worked for 31 days, and collected 168,610 records which contain 11,228 distinct publications. The results show that (1) with the highlight feature on popular items, the aggregate completion times of downloading resources yield to a long-tail distribution however deviating from the Zipf Law; (2) publications with attracting titles tend to spread much faster than others. It is indicated that the users’ selection of resources is strongly influenced by the portal web features in BitTorrent. An implication of the results is that the attackers can leverage these features and deteriorate the attack consequence significantly. |
|
Keywords:BitTorrent, User Behavior, Web Features, Content Pollution |
|