AI in Smart Grid Cybersecurity: A Systematic Review of Machine Learning and Deep Learning Approaches against False Data Injection and Other Emerging Attacks

Abstract

The increasing digitalization of smart grids has heightened their vulnerability to sophisticated cyber threats, with false data injection (FDI) and other emerging attacks posing significant risks to grid stability, reliability, and resilience. Artificial intelligence, particularly machine learning (ML) and deep learning (DL), has gained prominence as a promising defense layer capable of detecting, mitigating, and adapting to these dynamic threats. However, the rapid growth of research in this area has produced fragmented findings across diverse methodologies, datasets, and evaluation strategies. To address this gap, our systematic review consolidates the current state of ML- and DL-driven approaches in smart grid cybersecurity, with a specific emphasis on FDI detection and defense against evolving adversarial tactics. We map the landscape of proposed techniques, highlight benchmark datasets and simulation environments, and critically examine strengths, limitations, and open challenges. In doing so, we establish a taxonomy of AI-based solutions that organizes existing efforts by learning paradigm, attack type, and deployment layer within the smart grid. Beyond cataloguing current achievements, we underscore persistent challenges such as scalability, data imbalance, adversarial robustness, and model explainability, all of which constrain real-world deployment. By synthesizing insights from both academic research and industrial practice, this review aims to provide a roadmap for researchers, practitioners, and policymakers seeking to develop resilient, trustworthy, and adaptive AI-driven cybersecurity mechanisms for future power systems.