Adversarial Deep Neural Networks for Intrusion Prevention in Heterogeneous IoT Ecosystems

Over the past few years, the Internet of Things (IoT) technologies have grown to offer a large variety of heterogeneous devices, posing significant security threats to contemporary networked technologies. This chapter introduces a new adversarial deep neural network architecture that is explicitly created to provide a solid intrusion prevention service through a heterogeneous IoT ecosystem. Traditional intrusion detection systems based on machine learning were found to have a vulnerability coefficient of more than 0.73 to adversarial perturbation, which makes them ineffective against more advanced attacks. Generative Adversarial Networks and other deep learning concepts have become potent tools for generating data, strengthening their robustness and solving intricate problems in a variety of fields of use. Parallel deep learning pipelines have also been made more application-driven towards real-life sensing applications and also in medical imaging.