The exponential growth in data generated by interconnected devices within the Internet of Underwater Things presents exciting opportunities to improve the quality of service in emerging applications through enhanced data sharing. Underwater devices collect and transmit sensitive information like oceanographic data, marine life observations, or even information related to military activities. Ensuring privacy is crucial to protect this sensitive data from unauthorized access or misuse. Blockchain Technology provides security in sharing data. Federated Learning maintains privacy without sharing the actual data. In this article, we propose a Blockchain-Federated Learning for Privacy Preservation in the Internet of Underwater Things (BFL) which combines Blockchain Technology with Federated Learning to ensure the privacy of data. In this article, we use a consensus algorithm for secure data sharing and ResNet-50 to protect the privacy of data. We analyze the proposed approach with the existing approach and the result depicts that the proposed approach achieves \(\sim \) 16% better accuracy than the existing approach.