Online Social Networks (OSNs) play an important role for internet users to carry out their daily activities like content sharing, news reading, posting messages, product reviews and discussing events etc. At the same time, various kinds of spammers are also equally attracted towards these OSNs. These cyber criminals including sexual predators, online fraudsters, advertising campaigners, catfishes, and social bots etc. exploit the network of trust by various means especially by creating fake profiles to spread their content and carry out scams. All these malicious identities are very harmful for both the users as well as the service providers. From the OSN service provider point of view, fake profiles affect the overall reputation of the network in addition to the loss of bandwidth. To spot out these malicious users, huge manpower effort and more sophisticated automated methods are needed. In this paper, various types of OSN threat generators like compromised profiles, cloned profiles and online bots (spam bots, social bots, like bots and influential bots) have been classified. An attempt is made to present several categories of features that have been used to train classifiers in order to identify a fake profile. Different data crawling approaches along with some existing data sources for fake profile detection have been identified. A refresher on existing cyber laws to curb social media based cyber crimes with their limitations is also presented.