This paper presents automatic key generation for long-range wireless communications in low power wide area networks (LPWANs), employing LoRa as a case study. Differential quantization is adopted to extract a high level of randomness. Experiments conducted both in an outdoor urban environment and in an indoor environment demonstrate that this key generation technique is applicable for LPWANs, and shows that it is able to reliably generate secure keys.Index Terms-Internet of Things, low power wide area networks, physical layer security, key generation, LoRa/LoRaWAN
I. INTRODUCTIONThe Internet of Things (IoT) is capable of connecting people, things, and the environment. This revolution heavily relies on secure data communications, which are currently maintained by classic cryptographic algorithms and protocols. In particular, public key cryptography (PKC) has been the de facto scheme for distributing keys to the users in modern communication and computer networks. However, its application in the IoT remains a challenge owing to the limited computational and battery capacity, as well as the requirement of a public key infrastructure for distributing the public keys.Key generation from the wireless channel between any pair of users has become a promising design alternative to complement PKC. The keys generated can be used for the symmetric encryption schemes in different layers of the protocol stack, e.g., the Wi-Fi Protected Access (WPA) for the Wi-Fi MAC layer encryption or for Transport Layer Security (TLS) in the transport layer. It is particularly for protecting IoT systems that contain large numbers of resource-limited devices [1]. A comparison of resource and energy consumption between the key generation and elliptic curve-based Diffie-Hellman (ECDH) procedure, which is a popular PKC scheme, has been carried out in [2]. Key generation has been demonstrated to be more cost-efficient. Explicitly, ECDH consumes 98 times more energy and imposes 1289 times higher complexity than key generation, when both are implemented by an 8-bit Intel MCS-51 micro-controller [2]. In addition, key generation does not require any assistance from a third party, which is suitable for many decentralized or device-to-device IoT applications.