We propose Shapley Value based centrality measures for signed social networks. We also demonstrate that they lead to improved precision for the troll detection task.