Cyber-physical systems (CPS) have the great potential to transform people's lives. Smart cities, smart homes, robot assisted living, and intelligent transportation systems are examples of popular CPS systems and applications. It is an essential but challenging requirement to offer secure and trustworthy real-time feedback to CPS users using spectrum sharing wireless networks. This requirement can be satisfied using collaborative spectrum sensing technology of cognitive radio networks. Despite its promising benefits, collaborative spectrum sensing introduces new security threats especially internal attacks (i.e., attacks launched by internal nodes) that can degrade the efficiency of spectrum sensing. To tackle this challenge, we propose a new transferring reputation mechanism and dynamic game model-based secure collaborative spectrum sensing strategy (TRDG). More specifically, a location-aware transferring reputation mechanism is proposed to resolve the reputation loss problem caused by user mobility. Furthermore, a dynamic game-based recommendation incentive strategy is built to incentivize secondary users to provide honest information. The simulation experiments show that the TRDG enhances the accuracy of spectrum sensing and defends against the internal attacks effectively without relying on a central authority.