Peer-to-peer (P2P) is a newly emerging computing model succeeding conventional C/S model, and a peer in a P2P system may act as both a client and a server. Particularly, the P2P systems are with the absence of central control or global information. Since P2P environment is open and dynamic in nature, trust mechanisms modeling social networks are used to build trust in peers for safeguarding interactions of them. In this thesis, research on trust management in P2P systems is focused on, which involvles trust and reputaion model, trust-based application schemes, etc.In the thesis, a reputation management model is presented, in which degrees of trust and distrust with regard to a peer are evaluated based on its positive and negative behaviors, respectively. In the model, trust/distrust ratings for a transaction can be differentiated multiple levels on each aspect, and the more recent transaction experiences take up a more important role in the evaluation result. The model also introduces the Confidence Index (CI) for representing the confidence in evaluation results, so peers can communicate ratings for others to compute their reputations. Correspondingly, a provider selection scheme for file downloading is designed to isolate distrusted ones and promote trusted ones. As a result the performances in downloading are improved.Due to lack of sufficient knowledge in a P2P environment, it is difficult to express trust in other entities precisely, furthermore, trust evaluation usually depends on principal's subjective judgments, so fuzzy set based trust and reputation model is proposed for peers to build trust and reputation based on observations and recommendations. In the model, linguistic labels are used for presenting differentiated trust levels, and trust on different aspects in collaborations can be combined. In deducing peers' reputations, accuracy of recommendations is represented with fuzzy similarity measure, and then an algorithm is introduced to tune corresponding recommenders' impacts on theresults based on it. With this model, peers can easily select right collaborators, and form group sharing similar preferences.Since the effectiveness of any routing protocol is dependent on the P2P overlay network's interconnection topology, simultaneously controlling the network topology should improve the resource discovery performance as well. In the thesis, a trust-based P2P topology evolution mechanism is proposed, with which peers can build trust in others through interaction experiences or feedbacks from others, so those peers most likely to provdide satisfactory content are given to high trust. Peers can establish direct links to more trusted ones for facilitating interactions with them to get content wanted locally. With the establishment of links to more trusted ones and the removal of links to less trusted ones, peers' locally adaptive behaviors can lead to globally efficient P2P networks in resource discovery.This thesis also presents a trust manage framework middleware for P2P collaboration applications, the interfaces of which abstract the specific operation details, facilitate the customization for specific application extension, and the general framework services improve inter-operations between different collaboration applications. Trust-based collaboration decision service can provide collaboration decision service on top of different trust models, according to collaborative roles, trust, and context conditions. Automated negotiation framework services provide general P2P trust negotiation service interfaces, and define negotiation infrastructure and process.