Open Collections

Abstract

The rapid advancement of digital technologies has resulted in an increase in data generation and availability, along with breakthroughs in artificial intelligence, machine learning, and access to high-performance computing resources. These developments have drawn attention into data-driven approaches to tackle various challenges in industry. One of the most important challenges is enhancing security and protecting systems against malicious cyberattacks to prevent their catastrophic consequences. This thesis focuses on one of most common types of cyberattacks, called false data injection attack, which compromises the integrity of data transmitted over communication networks. This thesis aims to enhance the resiliency of networked control systems and ensure their safe operation in the presence of malicious activities through mitigating the impacts of false data injection attacks in real-time using a novel two-stage data-driven framework. The proposed framework utilizes a stacked ensemble learning architecture, including a variety of time series forecasting models, and a model selection module to preserve the computational efficiency of the developed framework. The first stage of the proposed framework involves meta learning to select a time series forecasting model from the stack for mitigation purposes. In the second stage, the selected model mitigates false data injection attacks in real-time. The proposed method's effectiveness is demonstrated through rigorous simulations involving the formation control of differential-drive mobile robots.