In recent years, social networks have experienced strong growth in both size and popularity. One of the main characteristics of these systems is their reliance on users as the primary contributors of content. This dependence makes the users of these systems the best targets for malicious behavior. In an effort to preserve community value and ensure long term success, the proposed approach is based on the use of social honeypots to discover malicious profiles in social networks. Inspired by security researchers who used honeypots to observe and analyze malicious activity in the networks, this method uses social honeypots to trap malicious users. The two key elements of the approach are: (1) the deployment of social honeypots for harvesting information of malicious profiles. (2) Analysis of the characteristics of these malicious profiles and those of deployed honeypots for creating classifiers that allow to filter the existing profiles and monitor the new profiles. In this paper, we describe the different steps of the proposed approach, starting with the deployment of social honeypots, the use of both feature based strategy and honeypot feature based strategy methods for collecting data, and finally the development of machine learning based classifiers for identifying malicious profiles.