In highly dynamic IoT environments, the connection statuses of IoT resources and the availability of IoT-based services change frequently. Therefore, to successfully build distributed service registries for managing and finding the information about available services in an effective manner, it is crucial to minimize the overhead of message exchanges between registries and to reduce the time overhead for identifying the capabilities of available IoT resources and the services that can be provided by utilizing these capabilities. In this paper, we propose a lightweight messaging approach that uses hierarchical Bloom filters to efficiently represent service information to be exchanged and managed by distributed service registries for IoT environments with high mobility. We also propose a method for serializing the dimensions of a Bloom-filter-encoded search space. We conducted experiments to demonstrate the improvement in the service discovery performance, the reduction in message traffic among service registries, and the decrease in the latency when synchronizing distributed service registries.