Manifold Learning-Assisted Physical Layer Key Generation for LoRaWAN: An Experimental Study

• Main Authors: Aghajari, Hossein, Bakhtiari Babadegani, Hamed, Naderi Soorki, Mehdi, Ahmadinabi, Sajad, Mohsen Ahmadi, Seyed
• Format: Conference Proceeding
• Language: English
• Subjects: Encoding; Information leakage; Logic gates; LoRa technology; Manifold learning; Manifolds; Physical layer; Physical Layer Security; Wireless communication; Wireless link key
• ISSN: 2642-9527
• DOI: 10.1109/ICEE59167.2023.10334776

Long Range Wide Area Network (LoRaWAN) has been widely proposed as one of the main promising access networks for the battery-constrained internet of things (IoT) device. Although LoRaWAN already provides many security features such as data confidentiality and integrity between LoRa end nodes (ENs) and application servers at the core, there is a lack of schemes to manage and distribute secure wireless keys between LoRa ENs and gateways at wireless access. In this paper, an efficient physical layer security-based scheme is proposed which explores the randomness of the received signal strength index (RSSI) of LoRa wireless signals to generate link key. Due to the resource constraints LoRa nodes, manifold learning methods are applied to reduce the dimensionality of measured data of channel vectors for initial key generation. Then, a bit disagreement in the initial keys between LoRa EN and gateway are detected and corrected by means of error correction coding. Finally, to prevent information leakage in the presence of attacked node, the cryptographic hashing algorithm is utilized to generate the final key from the initial keys. To analyze the performance of the proposed manifold learning-assisted physical layer key generation in real world, several experiments for different wireless LoRa links such as line-of-sight (LoS), non-LoS, and tree-covered areas are performed over the campus of Shahid Chamran University of Ahvaz. Our analysis of the experimental measurement shows that even when the attacker node is at 50 cm distance from the LoRa EN to recover the Link key, the bit disagreement rate between legitimate EN and attacker keys is 20% in all measurement scenarios. Moreover, we also find that the local tangent space alignment method for manifold learning leads to better security performance.