Recent years have witnessed an explosion of multimedia contents available online. However, noisy and spam tags provided by spammers dramatically degrade the trust level of Social Networking Service. In this paper, we propose a novel social multimedia tagging system based on a user trust model which exploits the low-level visual features of multimedia contents. Especially, we first analyze the low-level visual features of similar multimedia contents and then measure trust scores of each user. With the trust scores, we construct a new user trust model and propagate the most relevant and reliable tags to the untagged multimedia content based on the user trust model.