Abstract: Critical infrastructure systems, which were built decades before, such as power grids, water treatment systems, transport networks, industrial control systems, etc., are also still crucial to national security but more susceptible to advanced cyber attacks. The old-fashioned architectures, proprietary protocols, and patching possibilities render the traditional security approaches to be inadequate. The paper will discuss the potential of using AI-assisted DevSecOps to change the state of security of these old environments by embedding continuous monitoring, threat detectors, and secure-by-design within the entire system life cycle. We discuss recent innovations in machine learning-based anomaly detection, predictive maintenance, and automation in CI/CD pipelines, which have the potential to make them more resilient without affecting the continuity of the operation. The research suggests a hybrid AI-DevSecOps framework based on the current research and practical applications, allowing the implementation of this model in legacy systems. The framework focuses on safe configuration, model driven threat intelligence, auto compliance checking, and persistent checking. The findings indicate that a combination of AI and DevSecOps could help shorten the time to respond to the threat to a minimum, increase the visibility of the OT/IT boundaries, and allow proactive defense mechanisms. The paper finds that AI-assisted DevSecOps is a feasible and scalable solution that can be used to update the cybersecurity of legacy critical infrastructure without jeopardizing reliability, safety, and mission-critical performance.

Keywords: AI-enhanced security; DevSecOps; Legacy critical infrastructure; Industrial control systems; Machine learning; Cybersecurity automation; Anomaly detection

Downloads: | DOI: 10.17148/IJIREEICE.2025.13213