With the daily increasing demands for higher-quality services, mobile Internet of Things (MIoT) are facing various challenges, such as communication security, availability, scalability, etc. Its changing topology features bring a lot of trouble when solving the above problems. Though the evolved cellular network is expected to bring promising solutions, some inherent problems in traditional MIoTs may keep hindering its development. Thus, to overcome above problems, we propose a software-defined MIoTs-based model providing communication security and privacy protection under emergencies. In our scheme, the control plane is responsible to compute routes for online learning devices (OLDs), and forward entries for switches. Taking use of information that OLDs and facilities collect, controller is able to coordinate the overall situation. To ensure the authenticity and reliability of messages sent by OLDs, signature and authentication should be provided. We also introduce an emergency-dealing system. It transplants the multicast technology into software-defined MIoTs to generate a Steiner Tree among impacted nodes, so that OLDs will be informed as soon as emergency happening. The security analysis proves that our scheme is able to ensure the communication security in software-defined MIoTs. The evaluation of performance indicates that our scheme outperforms other existing schemes.