Scholarlink

Artificial Intelligence (AI) and Machine Learning (ML) play a major role in advancing cybersecurity anomaly detection, predictive maintenance, and human safety in industrial control systems (ICS) in the contemporary era. In the context of cybersecurity for ICS, anomaly detection is one of the most prominent areas of research. Most research focuses on adopting different algorithms and fine-tuning methods to improve model performance, detection capabilities, and prediction accuracy. However, less attention has been given to incorporating security best practices during the development of AI agents or ML models. In this paper, we review existing research on AI- and ML-based security anomaly detection in ICS to analyze the extent to which cybersecurity best practices are considered. We specifically examine how challenges such as data poisoning, privacy attacks, and evasion attacks are addressed during the development of AI and ML models for ICS. Based on our findings, we propose best practices that can be adopted to enhance the security and resilience of AI and ML models in industrial control system environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Review of Research Work on Cybersecurity Best Practices Adopted During AI and ML Development for Industrial Control Systems

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Suresh Kumar Kasi,
  • Manju Bargavi Sankari Krishnamoorthy

摘要

Artificial Intelligence (AI) and Machine Learning (ML) play a major role in advancing cybersecurity anomaly detection, predictive maintenance, and human safety in industrial control systems (ICS) in the contemporary era. In the context of cybersecurity for ICS, anomaly detection is one of the most prominent areas of research. Most research focuses on adopting different algorithms and fine-tuning methods to improve model performance, detection capabilities, and prediction accuracy. However, less attention has been given to incorporating security best practices during the development of AI agents or ML models. In this paper, we review existing research on AI- and ML-based security anomaly detection in ICS to analyze the extent to which cybersecurity best practices are considered. We specifically examine how challenges such as data poisoning, privacy attacks, and evasion attacks are addressed during the development of AI and ML models for ICS. Based on our findings, we propose best practices that can be adopted to enhance the security and resilience of AI and ML models in industrial control system environments.