Underwater wireless sensor networks (UWSNs) are dynamic and open systems that present substantial problems to secure data transfer because malevolent nodes can intercept and tamper with sensitive data. This study suggests an Encoder-Based Trusted Routing Algorithm (ETR) for UWSNs in order to overcome these security concerns. In contrast to current approaches that depend on Generative Adversarial Networks (GANs) for assessing trust. The suggested technique makes use of an encoder to create a trust profile for every network node according to attributes including position, signal intensity, and energy usage. Even in the presence of unlabeled and imbalanced training data, the encoder is able to distinguish between trusted and malicious nodes with high accuracy by utilizing an unsupervised learning technique.