Proactive Defense for Fog-to-Things Critical Infrastructure

Abstract

Robust and adaptable cybersecurity mechanisms are needed to mitigate sophisticated and future zero-day cyberattacks and threats, particularly in the dynamic Fog of Things (FoT) computational paradigm, which makes use of massively distributed nodes. Deep learning (DL)-driven architectures have been proven more successful in big data areas than classical machine learning (ML)-based algorithms. We orchestrate the software defined networking (SDN) control plane to propose a highly scalable proactive defense mechanism leveraging the Cuda-Deep Neural Network Gated Recurrent Unit (CU-DNNGRU) for the FoT critical computing infrastructure. Furthermore, the proposed framework does not place an extra burden on the underlying energy- and power-constrained FoT devices. We used the current state-of-the-art dataset (i.e., CICIDS2018) and evaluated our approach using standard performance metrics. We compare our proposed technique with our constructed hybrid DL-driven architectures and benchmark DL algorithms to evaluate its performance and efficacy. We hope that this work will enable further security research in the next-generation FoT computational paradigms.