Cooperation among wireless nodes has been recently proposed for improving the physical layer (PHY) security of wireless transmission in the presence of multiple eavesdroppers. While existing PHY security literature answered the question “what are the link-level secrecy rate gains from cooperation?”, this paper attempts to answer the question of “how to achieve those gains in a practical decentralized wireless network and in the presence of a cost for information exchange?”. For this purpose, we model the PHY security cooperation problem as a coalitional game with non-transferable utility and propose a distributed algorithm for coalition formation. Through the proposed algorithm, the wireless users can cooperate and self-organize into disjoint independent coalitions, while maximizing their secrecy rate taking into account the security costs during information exchange.We analyze the resulting coalitional structures for both decode-and-forward and amplify-and-forward cooperation and study how the users can adapt the network topology to environmental changes such as mobility. Through simulations, we assess the performance of the proposed algorithm and show that, by coalition formation using decode-and-forward, the average secrecy rate per user is increased of up to 25.3% and 24.4% (for a network with 45 users) relative to the non-cooperative and amplifyand-forward cases, respectively.