Online social networks allow millions of individuals to create online profiles and share information with vast networks of friends, and often, unknown strangers. Privacy within social networking sites is often undefined, which might render potential privacy risks. In this paper, we present a dynamic trust-based privacy assignment system to help people select the privacy preference on-the-fly to the piece of content he/she is sharing, where trust information is derived from social network structure and user interactions. Our proposed system, Personal Social Screen (PerCial), first automatically detects a two-level topic sensitive community hierarchy using the available resources, and then assigns privacy preference for users based on their personalized trust networks. Preliminary results on a social object network dataset collected from Flickr demonstrate the efficacy and effectiveness of our proposed system.