For the large scale of the network and unreliable nature of peers in most P2P (Peer-to-Peer) systems, it is a challenging task to implement an efficient and robust P2P reputation system. Self-storage P2P reputation systems provide an efficient way to retrieve rating values. Since each peer locally stores its own rating values given by all other peers, it takes only one query message to collect all the rating values of a peer. However, it also gives the owner control over its rating values. One of the challenges of this type of system is to protect the rating values from being maliciously modified by their owners. In existing self-storage P2P reputation systems, other participants are involved in order to ensure the integrity and reliability of the rating values. However, because at least one participant must be contacted during the rating and rating values retrieval processes, they create a high message overhead when deployed in unstructured P2P systems.

In this article, we address the efficiency and robustness of P2P reputation systems in fully distributed environments. We propose an innovative self-storage P2P reputation system. In our system, each peer locally stores its own positive rating values given and signed by other peers. On one hand, since it is not likely for a peer to discard any part of the positive rating values or forge the raters' signatures, the integrity and reliability of the self-storage rating values are preserved. On the other hand, based on the self-storage rating values and our reputation value computing algorithm, one step is enough to rate or obtain the whole positive rating values of a peer without other participants or broadcasting messages. Extensive simulations have been conducted to evaluate the performance of our system. Simulation results show that the message overhead of our system is reduced significantly (4 messages for each transaction) compared with existing self-storage P2P reputation systems, which makes our system a scalable reputation system for handling a large number of P2P services. Our system is also robust to malicious behaviors of peers. Simulation results show that the service selection success ratio is greater than 0.8 even when there are 70% malicious peers. Moreover, when those participants are not reachable in existing self-storage P2P reputation systems, the rating values verification process fails, while in our system, no other peer is involved and this type of threat can be totally avoided.