A Mutual Security Authentication Method for RFID-PUF Circuit Based on Deep Learning

The Industrial Internet of Things (IIoT) is designed to refine and optimize the process controls, thereby leveraging improvements in economic benefits, such as efficiency arid productivity. However, the Radio Frequency Identification (RFID) technology in an IIoT environment has problems such as low security and high cost. To overcome such issues, a mutual authentication scheme that is suitable for RFID systems, wherein techniques in Deep Learning (DL) are incorporated onto the Arbiter Physical Unclonable Function (APUF) for the secured access authentication of the IC circuits on the IoT, is proposed. The design applies the APUF-MPUF mutual authentication structure obtained by DL to generate essential real-time authentication information, thereby taking advantage of the feature that the tag in the PUF circuit structure does not need to store any essential information and resolving the problem of key storage. The proposed scheme also uses a bitwise comparison method, which hides the PUF response information and effectively reduces the resource overhead of the system during the verification process, to verify the correctness of the two strings. Security analysis demonstrates that the proposed scheme has high robustness and security against different conventional attack methods, and the storage and communication costs are 95.7% and 42.0% lower than the existing schemes, respectively.