A deep learning-based authentication protocol for IoT-enabled LTE systems

Abstract

The connected devices in Internet-of-Things (IoT)-enabled systems are continuously increasing nowadays, and likely to grow exponentially worldwide in near future. Hence, the next generation IoT-enabled mobile networks (e.g., 5G onward) are expected to provide higher system capacity and ultra-low latency to deal with. According to the Third Generation Partnership Project (3GPP), Long-Term Evolution (LTE) technology can serve the purpose efficiently, and also bridge the gap between earlier and future generation mobile networks. However, the network may face problems associated with privacy and security, as the underlying communication is mostly wireless. Thus, a secure and efficient Authentication and Key Agreement (AKA) protocol is desirable. Recently, many protocols have been proposed to address these goals. Unfortunately, the security and efficiency of such protocols are still in doubt. This paper introduces a deep learning-based AKA protocol for IoT-enabled LTE systems. The proposed protocol can address mutual authentication among the communicating entities. It employs a Deep Residual Network (DRN)-based key generation technique, called DRN-KeyGen, to establish a shared secret key on-the-fly among the communicating entities in such environment where the number of IoT devices are excessively large and extremely heterogeneous in nature. The security of DRN-KeyGen is verified considering various active and passive attacks that may occur in wireless-enabled LTE systems. The efficiency of DRN-KeyGen is measured through two different parameters: attack detection rate and attack detection time, where each attack is experimented using Python tool-based simulation with varied key lengths. The empirical results show that the proposed DRN-KeyGen can achieve an average detection rate of 0.924 and an average detection time of 30.634 s considering key lengths of 32 bits, 64 bits, 128 bits, 192 bits, 256 bits, and 512 bits. Finally, we have compared the security and efficiency of DRN-KeyGen with other existing protocols to show its superiority.