Hierarchical Deep Learning for Cybersecurity of Critical Service Systems
Lav Gupta · 2020 Fourth World Conference on Smart Trends in Systems, Security and Sustainability (WorldS4) · 2020
The costs of delivering critical services, such as healthcare, power, financial systems and transportation are spiraling up while the performance expectations of them have risen manifold. The mechanics of providing these services (e.g., activities like processing satellite imagery for defense and civil applications, analyzing patient data for diagnosing acute ailments, ensuring uneventful working of unmanned vehicles and big data analytics for sustaining smart cities) all require unprecedented data acquisition, storage, computation and communication. Driven by the need to achieve agility, intelligence and high performance at lower cost and improve outcomes, these systems are increasingly relying on advanced methods and technologies. Trends show the prevalence of IoT for data collection, multi-cloud computing for storage and analytics and virtualization of communication components. The increasing sophistication of technology also increases susceptibility of these systems to cyberattacks. The overarching objective of the work presented in this paper is to show the feasibility and usefulness of AI-based hierarchical sparse deep neural network models for providing security to data in motion. The novelty of the work lies in the use of variable complexity models across multiple clouds and their synergistic training to reduce training time and improve detection accuracy for unknown attacks.