In recent years, due to the proliferation of radio fre-quency identification (RFID) technology in everyday life, espe-cially in critical domains such as health care and transportation systems, significant efforts have been made to enhance the de-pendability of this technology. From these efforts have emerged specific techniques and several middleware solutions to handle the large amount of the RFID data. These solutions are not suita-ble for all RFID system requirements especially for issues related to critical domains. In this paper, we propose a novel fault-tolerant RFID middleware providing two fault-tolerant mechan-isms. The first mechanism is an online diagnosis algorithm based on a statistical analysis of the generated RFID data to identify faulty components of the system such as faulty readers or tags. The second mechanism is a verification process based on an extended finite state machine of the Low Level Reader Protocol (LLRP), the communication standard between RFID readers and RFID middleware. This process aims at identifying the causes of the diagnosed failures.