Texas A&M Researchers Develop Cyber Defense System for Critical Infrastructure

Researchers have developed RADIANT, a new cybersecurity system that enhances protection for critical infrastructure by detecting stealth cyberattacks without the need for costly retraining.

Oct. 22, 2025

2 min read

Researchers from Texas A&M University’s CARES Laboratory have developed RADIANT, a cybersecurity system designed to detect stealth cyberattacks on critical infrastructure such as power grids and water systems without requiring costly retraining.

A research team from Texas A&M University’s Clean and Resilient Energy Systems (CARES) Laboratory has introduced a new cybersecurity system designed to detect and defend against stealth cyberattacks targeting critical infrastructure such as power grids and water systems.

The system, called the Reactive Autoencoder Defense for Industrial Adversarial Network Threats (RADIANT), was developed under the leadership of Dr. Irfan Khan. It identifies advanced adversarial threats without requiring the costly retraining typically needed by traditional intrusion detection systems.

Stealth cyberattacks represent a sophisticated class of adversarial threats that disguise malicious activity as legitimate network traffic, making them difficult for both automated detectors and human operators to identify. RADIANT functions as a reactive defense layer that complements existing cybersecurity systems, improving detection capabilities while avoiding full system retraining.

According to Khan, RADIANT was designed to maintain reliable detection and operator confidence even when attackers attempt to induce machine-learning systems to misclassify malicious activity as normal. The system works by reconstructing incoming data and identifying inconsistencies, filtering out manipulations to enhance accuracy and reduce false alarms.

Lead author Syed Wali Abbas Rizvi, a Ph.D. student in the Department of Electrical and Computer Engineering and researcher in the CARES Laboratory, described RADIANT as “deployment oriented,” emphasizing its ability to integrate with existing intrusion detection systems in substations, microgrids and process plants with minimal overhead.

The researchers plan to expand RADIANT’s capabilities by testing against adaptive adversaries aware of its methods and exploring broader types of decision-based attacks. Future work also includes operator-in-the-loop field studies to measure performance in real-time plant environments.

Also contributing to the project was Ph.D. student Yasir Ali Farrukh, with funding support from the Office of Naval Research.