With the widespread adoption of distributed energy resources, the authentication of a large number of distributed "edge-to-end" devices and the identification of their behaviors within active distribution substations are becoming increasingly challenging. The damaged devices may jeopardize the security of the distribution network, while the malicious devices can disrupt the integrity of the network by injecting false and malicious data. Compared to high-cost key-based security solutions, employing trust-based security to detect malicious nodes is an effective and lightweight countermeasure. This paper proposes an efficient trust evaluation mechanism that can effectively distinguish malicious devices and defend against switch and Denial-of-Service(DoS) attacks. Using the Bayesian estimation method, the direct and indirect trust values of edge-end devices are collected and calculated, further considering the correlation of data collected over time. By applying a two-stage trust evaluation framework, accurate trust evaluation is achieved in dynamic environments, thereby ensuring the security of data transmission. Compared with existing methods, the proposed method has less delay in detecting malicious nodes and higher network throughput.