Discrimination of Zombie Fans on Weibo based on Features Extraction and Business-Driven Analysis

Weibo has attracted, due to its popularity and openness, massive zombie fans to achieve their some certain purpose, to follow somebody or repost with spam content. Zombie fans are generally produced and maintained by automated programs. They seriously interfere in normal social environment and make human users experience deteriorates. We propose a new approach that is able to deal with evolutions and variations of zombie fans. Compared with the traditional detecting methods, the new approach is innovative in two respects: (a) we observe the difference and extract the features between human and zombies in terms of their behavior, message content, and profile properties; (b) we introduce a new concept of business-driven analysis why somebody tend to purchase zombie fans, based on which a new discrimination way is delivered. We conduct large-scale measurements with a collection of over 50,000 user accounts from Weibo. The experimental evaluation demonstrates the efficacy and stability of the proposed classification system.