Cyber Attacks Detection based on Deep Learning for Cloud-Dew Computing in Automotive IoT Applications

The Internet of Things (IoT) technology is exponentially becoming more relied on in different industrial disciplines including automotive mobility applications. It is an integral component of the transportation mobility with the vision to acquire and process data for improving traffic safety. However, it has not matured, yet, considering the threats of network intrusion, and the level of autonomy the technology is aiming to achieve. This is a critical factor in terms of the user's physical and cyber safety. In this paper, we evaluate the case of transferring data between the cloud and the enduser dew devices integrated into the connected vehicle. We analyze the implementation and organizational approaches related to Dew Computing, where the processing is brought even closer to the user compared to other IoT computing paradigms. This paper aims to present a threat analysis of the IoT and to use a deep learning approach to counter cyber anomalies, then validate it by analyzing its metrics. We use a modified version of the Stacked Autoencoder that improves the accuracy of detecting the defined attacks, using the loss over the training data as a threshold. Our approach gives an ameliorated outcome, matching 90% accuracy, compared to previous models.