Distributed Hash Table (DHT) network is widely used in secure self-destruction schemes of privacy data in cloud computing environment, but malicious nodes and dishonest nodes in the DHT network easily lead to key shares loss or leakage. To tackle those problems, a trust evaluation mechanism was proposed for the DHT network used in cloud-data secure self-destruction system. In this mechanism, a trust cloud model was established for DHT nodes to describe their trust information qualitatively and quantitatively. By introducing an improved calculation method of direct trust value together with recommended trust value and fully considering the internal and external factors of DHT nodes, the trust value of nodes were first calculated on two dimensions consisted of operating experiment and interactive experience. The result data were used to build trust evaluation sub-cloud for each index. After that, all these trust evaluation sub-clouds were summed up to generate the comprehensive trust cloud according to the weights of different evaluation indexes. Then, the comprehensive trust cloud, by means of cloud generator algorithm, could be described as one-dimensional normal cloud. Finally, the reliable and efficient nodes could be selected using trust decision algorithm. Experimental results show that the proposed mechanism can help original data self-destruction system making comprehensive trust decision and finding reliable DHT network nodes, further enhancing disaster-tolerant capability and reducing computational cost of the system.