A trust-based P2P resource search method integrating with Q-learning for future Internet

Since Peer-to-Peer technology plays a vital role in the future Internet, such as the P2P video service Voddler, music service Spotify and Symform are designed for cloud storage aroused users’ high interest. The P2P system has been attracting more and more attention on resource location and service discovery. To find the desired resources and achieve file sharing more effectively and smoothly, we propose a novel P2P search method integrating trust mechanism with Q-learning method (SMITQ) for future Internet in this paper. Q-learning, which is one of the most important reinforcement learning models with the self-adaption and feedback learning mechanism, facilitates P2P searching efficiency and would be widely used in routing-aware for future Internet. By introducing node trust mechanism, we construct neighbor nodes evaluation matrix to calculate the integrated information entropy of trust values and Q-values. The neighbor node with maximum integrated information entropy is selected as next hop to forward query message, which can avoid searching process with low efficiency due to the lack of information in Q table at initial search stage. The SMITQ method can speed up Q-learning process and provide an efficient method to select the neighbor node with the query resource. Simulation results indicate that the proposed method can obtain the highest success rate with the fewest duplicated messages as well as the shortest response time comparing with the Q-learning method and random walk method.