Rules are key elements for the collaborative interaction of different devices in IoT systems. However, the rule conflict problem is getting worse as IoT system complexity increases. And it is challenging to formally specify rules and detect rule conflicts for IoT systems with continuous states and running under uncertainty environment. To this end, this paper proposed a rule modeling and conflict detection approach called CredRule, aiming at construct “credible rules” for IoT systems. Firstly, CredRule provided a detailed definition for IoT rules, and proposed to specify rules, continuous system states and uncertainty environment with the NPTA (Network of Priced Timed Automata for short) model. Secondly, it provided a rule conflict taxonomy, and established a Rule Conflict Interaction Analysis model, providing an explicit description on rule conflicts. Finally, on basis of the established two models, it provided an automatic rule detection and analysis method based on the statistical model checking technique. To validate the effectiveness of CredRule, we carried out a case study on one ventilation and air conditioning system. The results demonstrate that CredRule can formally specify rules and effectively detect rule conflicts under uncertainty and continuous environment.