Generating hash value at edge device using SRAM-PUF and autoencoder in IoT network

With the deployment of the Internet of Things (IoT) networks, numerous IoT devices are placed in remote and often untrusted locations, such as open or wild environments. The reliability of such systems heavily depends on the data signals transmitted by these IoT devices, making it crucial to authenticate their trustworthiness. Current methods typically use cryptographic approaches to create encrypted digital signatures stored in volatile or non-volatile memory of the devices. However, these techniques are vulnerable to side-channel and replay attacks and incur significant energy and computational overheads. This paper proposes a novel approach for generating a device-specific identifier, or hash value, for each edge device. This hash value is derived from physically unclonable functions (PUFs), which utilizes manufacturing variations to create a unique device fingerprint. We design a system structure with edge devices that regularly authenticate all edge devices in the network using challenge-response pairs (CRPs). The authentication process employs PUFs and a Convolutional Neural Network (CNN) autoencoder to generate hash values for each device. Our experimental results demonstrate that the proposed scheme is highly efficient in authenticating and recognizing resource-constrained edge devices, with low computational overheads being a significant advantage of our approach.