In today's world, most of the people are using social networks for day-to-day activities. The most frequently used social sites are Facebook, Twitter, Google+, etc. These popular social networks are used by some of the users for abnormal or illegal activities. It is very important and necessary to identify and avoid such illegal activities without harming anyone in the society. In recent decades, social networks are becoming a popular research area for most researchers. Many authors are doing research on social network datasets and proposing various anomaly detection mechanisms to identify anomalous activities in both static and dynamic growing social networks. Various anomaly detection techniques are proposed by the authors to investigate malicious activities in social networks. In general, the process of identifying anomaly activities of the users in the given dataset is called anomaly detection. The anomaly detection in social networks is the process of investigating whether the users of the given social networks are involved in illegal activities or not. In this work, we proposed a most elegant approach to identify the anomalous or outlier users in the given social network. The proposed approach is considering the users participated in multiple communities of social networks. The designed algorithms are implemented and tested in a big data environment three node cluster using open source Hadoop ecosystem tools. Algorithm1 is used to investigate the nodes/users who participated in multiple communities of the given social network’s dataset. Algorithm2 takes the set of users participated in multiple communities and apply graph metrics such as degree and community score to predict the users involved in the anomalous activity.