Trust management frameworks are used to evaluate and manage trust relationships between network nodes and enhance network security. However, trust management frameworks themselves are vulnerable to attacks. Attacks against trust management frameworks are described in this paper with a trust management framework to resist them. The trustworthiness between nodes is evaluated to classify node behavior using a three-dimensional classifier based on a fuzzy integral. Different behaviors are mapped to different behavioral spaces to detect malicious nodes and identify their behavior types. The security of ad hoc networks is then improved by various measures to handle different types of malicious behavior. Simulations of the model on the System In The Loop (SITL) platform show that this trust management framework can separate normal nodes and malicious nodes and can distinguish different types of malicious nodes.