Few-Shot Learning for CPS Anomaly Detection: A Survey on Smart Grid Applications
Chowdhury Tasnuva Hazera, Mohamed I. Ibrahem, Mostafa M. Fouda · 2025
Cyber-Physical Systems (CPS) integration into critical infrastructures like smart grids and industrial control systems at such rapid speed leads to important vulnerabilities, while increased automation and efficiency are achieved. Traditional intrusion detection methods, however, face challenges associated with dynamic environments, lack of data, and evolving attack patterns. Few-shot learning (FSL) is an emerging solution where models can detect and classify certain classes with high accuracy using limited labeled data. This survey discusses applying FSL techniques, including Siamese Neural Networks and prototypebased models, for anomaly detection in CPSs, focusing on smart grid applications. We analyze recent developments, highlighting their strengths and limitations, and identify key opportunities for enhancing smart grid security. The present survey underlines how FSL is poised to address the new challenges to CPS while creating a road map for developing adaptive and resilient cybersecurity solutions, thus effectively bridging theoretical developments with practical challenges. Index Terms-Cyber-physical systems, fewshot learning, smart grid, intrusion detection