Bit flipping attack detection in low power wide area networks using a deep learning approach

Low Power Wide Area Networks (LPWANs) are a class of wireless technologies with characteristics such as large coverage areas, low bandwidth, very small payload size, and long battery life operation. LPWANs applications are growing rapidly in the industry, although their restricted resources have created some vulnerabilities and attacks. Bit flipping attack is such a potential attack in encrypted LPWANs data in which an attacker, with the knowledge of the bit location of each data field, can alter data without needing decryption of the message. Current methods of bit flip detection are based on the use of traditional integrity checking methods such as digital signature, message integrity code, or some sender-side methods that either increase the sender’s power consumption or the size of the message payload, hence are not practical in LPWANs. This paper proposed a novel bit flipping detection method through machine learning which eliminates the need to increase neither the payload size nor the sender’s power consumption. The proposed method is based on pattern recognition of true messages sequence using a deep neural network which is then capable to detect tampered messages accordingly. The performance of the method has been evaluated in some case studies and experiments including different modes of attacks, applications, and environment and the results show up to 99.84% accuracy in bit flipping attack detection.