Crowdsourcing has emerged as an important data collection paradigm in participatory and human-centric sensing applications. While many crowdsourcing studies focus on sensing and recovering the status of the physical world, this paper investigates the problem of profiling the crowd sensors (i.e., humans). In particular, we study the problem of accurately inferring the home locations of people from the noisy and sparse crowdsourcing data they contribute. In this study, we propose a semi-supervised framework, Where Are You From (WAYF), to accurately infer the home locations of people by explicitly exploring the localness of people and the dependency between people based on their check-in behaviors under a rigorous analytical framework. We perform extensive experiments to evaluate the performance of our scheme and compared it to the state-of-the-art techniques using three real world data traces collected from Foursquare. The results showed the effectiveness of our scheme in accurately profiling the home locations of people.