There is a need for efficient communication protocols that allow the private and scalable deployment of RFID systems with a large number of tags. In this paper, we leverage the idea of using distributed, collaborative readers to identify RFID tags and propose a new probabilistic communication protocol for those readers to privately identify RFID tags more efficiently in terms of computational cost and bandwidth usage. Our protocol, which is based on Hash-locks, allows readers to exchange information so as to reduce the amount of tag IDs stored in their caches. Consequently, our proposal improves the scalability of the system, and allows the easy management of large amounts of tags. We provide simulation results showing that our proposal is more efficient and flexible than previous ones in terms of computational cost and bandwidth usage.