Detection of malicious social bots: A survey and a refined taxonomy

Abstract

Social hots represent a new generation of hots that make use of online social networks (OSNs) as command and control (C&C) channels. Malicious social hots have been used as tools for launching large-scale spam campaigns, promoting low-cap stocks, manipulating users' digital influence, and conducting political astroturfing. Recent studies in this area either focus only on general security issues related to social networks or on coarse-grained categorization to support detection approaches. This survey aims to provide a comprehensive analysis from a social network perspective. To this end, we first categorize social bot attacks at different stages, then provide an overview of different types of social hots. Next, we propose a refined taxonomy that shows how different techniques within a category are related or differ from each other, followed by a detailed discussion of the strengths and limitations of each method. Following this, we review the existing datasets and summarize the results of empirical investigations. Finally, we highlight the limitations of existing detection approaches and suggest future directions for further improvement. Our study should help OSN administrators and researchers understand the destructive potential of malicious social hots and improve upon the current defensive strategies.