Cyberbullying has attracted the increasing attention among researchers. Social and computer science researchers have explored cyberbullying from various perspectives. This paper surveys the existing work on cyberbullying detection in social and computer science domains. It first introduces the basic research problems and characteristics of cyberbullying; second

it discusses a variety of machine learning algorithms for cyberbullying detection

including supervised learning

weakly supervised learning

rule-based and deep learning algorithms; and third

it summarizes 12 existing datasets used in cyberbullying detection and the popular metrics for detection performance. Finally

the paper analyzes the potential research from several aspects

such as cyberbullying detection approaches based on heterogeneous information network

auxiliary information fusion

and psychological characteristics.