When transmitting JSON data from an Edge AI device via MQTT to a central server, how can I optimize the message structure to minimize latency on low-bandwidth networks? Also, what are the best security practices for integrating HTTPS/REST API calls to a language model in a corporate environment?