Tagging in online social networks is very popular these days, as it facilitates search and retrieval of diverse resources available online. However, noisy and spam annotations often make it difficult to perform an efficient search. Users may make mistakes in tagging and irrelevant tags and resources may be maliciously added for advertisement or self-promotion. Since filtering spam annotations and spammers is time-consuming if it is done manually, machine learning approaches can be employed to facilitate this process. In this paper, we propose and analyze a set of distinct features based on user behavior in tagging and tags popularity to distinguish between legitimate users and spammers. The effectiveness of the proposed features is demonstrated through a set of experiments on a dataset of social bookmarks.