Abstract
Social networks are rapidly becoming part of our everyday activities. In online social network (OSN) environment, there is a huge amount of information which is available and widely used for various areas; such as provide the sharing of information and create relationship between people in a virtual community, capturing the criminals, detect terrorist and unlawful activities. Based on analyzing the OSN, there are two types of data that are inferred, first is behavioral data which depends on the dynamic behaviors of the user, and second is structural data which includes network structure. In social networking, there are enormous of anomalies. For instance; identity theft, hack account, fake account, spams and many other illegitimate activities, for this reason, there is a need for a way to detect these anomalies. There are many studies that conducted to detect the anomaly, but to the best of our knowledge, there were very limited researches carried out in the graph anomaly detection. However, those researches which used various data mining approaches are not promising, due to time complexity, lack of datasets, and lower accuracy. This paper attempts to present and discuss the previous works proposed to detect the anomalies on the OSN