A Pilot Study and Survey on Methods for Anomaly Detection in Online Social Networks

Abstract

An anomaly in online social networks is an unusual occurrence that deviates greatly from the standard. Financial fraud, social spam, and network intrusion can all be prevented with anomaly detection, which is designed to find this infrequent observational data. Anomaly detection on social media is crucial for preventing hostile actions like bullying, the spread of fraud information, etc. New forms of aberrant behavior have come to the forefront due to the increasing popularity of social media, creating alarm among a variety of groups. While classic anomaly detection problems have received a lot of attention, we have noticed a rise in research attention in the emerging area of social media anomaly detection. We classify anomalies as static versus dynamic, attributed versus unattributed, and investigate methods for identifying various forms of anomalies. We also analyze the benefits and disadvantages of various methodologies in each category, as well as the problems that this study area presents. Finally, we present challenges for detecting graph anomalies.