A Practical Side-Channel Attack of a LoRaWAN Module Using Deep Learning

With the development of Internet-of-Things technology, the electronic devices nowadays are embedded with “smart” functionalities, which enable them to upload data to the cloud for the easy access of the users. Such functionalities have been popular among commercial products. However, as we shall show in this paper, these functionalities, if not protected properly, can be vulnerable to attacks. An unprotected implementation leaks significant level of information related to the sensitive operations, which could pose threats to the security of the device. In this paper, we demonstrate a practical attack against a LoRaWAN communication module. This attack uses the deep-learning-based side-channel attack to recover the key for encrypting the payload data. By capturing less than 100 communication packets, the trained convolutional neural network is capable of recovering the full key for AES encryption.