Cold chain logistics is of great importance for transporting temperature and vibration sensitive products. However, fine-grained surveillance remains challenging in cold chain logistics, due to the lack of multi-dimensional information that reflects the status of monitored objects. In this paper, we propose a multi-dimensional information sensing surveillance framework, named MISS, to timely detect abnormal events that occur in cold chain logistics. The sensed information, including temperature and acceleration etc., can be integrated to provide accurate detection on the abnormal events. By adopting minimum entropy and AVC algorithms, we can classify various status in cold chain logistics. We further perform real implementations and evaluations on a prototype, and examine the effectiveness of MISS.