Abstract: With the rapid development of the internet of things (IoT), the number of IoT and edge devices continues to grow, and the associated security risks are becoming increasingly prominent. While traditional measures for transmission and storage security are relatively mature, new threats such as data tampering and privacy leakage still persist in the sensing and computing processes. To this end, this article systematically analyzes the core challenges of confidential sensing in IoT, including the difficulty of constructing trustworthy drivers for resource-constrained devices, the trade-off between privacy protection and efficiency in on-device model inference, and the inefficiency of securing complex IoT applications. In response to these challenges, this article further investigates feasible technical approaches such as automated construction of trusted I/O drivers, secure and efficient on-device inference, and automated security detection and hardening of IoT applications. Lastly, this article discusses future trends, including lightweight trusted execution environments, trustworthy IoT system foundations, and the evolution of distributed confidential computing and confidential interconnection.