A Deep Learning and Channel Sounding Based Data Authentication and QoS Enhancement Mechanism for Massive IoT Networks

The Internet of things has emerged as a technology that is affecting a lot of domains such as manufacturing and automation, smart traffic systems, security, disaster management, etc. Security and user authentication are challenging due to the large number of connected devices and the magnitude of data shared among the devices. Typically, some digital fingerprint in terms of the features of the data stream to be transmitted is embedded in the data streams, but they can be extracted in case the adversary analyses the data stream and records it for a long period with a sufficient number of samples. Moreover, large length stochastic features would inevitably increase the system computation overhead and latency at the gateway. While lesser overhead can be settled that would result in higher bit errors and chances of attacks. In this paper, a deep learning-based approach is used to detect possible attacks based on the statistical features embedded into the bitstream transmitted. Additionally, the channel state information has been utilized for enhancing the Quality of Service of the system. The performance metrics are the bit error rate, number of epochs for training, and mean square error of the deep learning model.