October 16, 2025
The GIST A RADIANT future for cybersecurity 
Sadie Harley
scientific editor
Andrew Zinin
lead editor
Editors' notes
This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:
fact-checked
trusted source
proofread
Stealth cyber-attacks against power grids, water systems and other critical infrastructure often go undetected until it is too late. Led by Dr. Irfan Khan, researchers from the Clean and Resilient Energy Systems (CARES) Laboratory at Texas A&M University have developed a novel cybersecurity system to detect and defend against these attacks.
In a paper published in Computers & Security, researchers introduce the Reactive Autoencoder Defense for Industrial Adversarial Network Threats (RADIANT), a new intrusion detection system that mitigates adversarial threats without relying on retraining.
When trying to attack an industrial control system of critical infrastructure, an adversarial cyber-attacker will deliberately craft an attack that causes the detection system to assign benign labels to malicious activity.
RADIANT's goal is to prevent stealth attacks—an advanced subclass of adversarial attacks that camouflage malicious activity as legitimate network traffic. During a stealth attack, both automated threat detectors and human operators may be shown seemingly "normal" signals while control objectives are being compromised.
Current intrusion detection systems require systems to be entirely retrained to mitigate stealth attacks, which can be ineffective against future threats and costly due to the required resources. RADIANT acts as a reactive defense layer that works with existing cybersecurity systems to detect suspicious activity, preventing costly retraining and increased defense against future attacks.
"We addressed the challenge of sustaining reliable detection and operator confidence when adversarial attacks—particularly advanced, stealth variants that mimic benign behavior—induce machine-learning–based intrusion detection systems to misclassify malicious activity as normal," said Khan, an affiliated faculty member in computer science and engineering and electrical and computer engineering, and an assistant professor of marine engineering technology in Galveston.
"Our objective is to increase robustness under attack while preserving accuracy on nominal operations, and to achieve this without continual adversarial retraining, enabling practical deployment in time-critical industrial environments."
RADIANT works by reconstructing incoming data and checking for inconsistencies—flagging suspicious cases for further inspection. This method filters out adversarial manipulations and improves detection accuracy while minimizing false alarms.
Researchers plan to continue making improvements to RADIANT by extending testing to adaptive adversaries with knowledge of the system's methodology, and broader families of decision-based attacks.
The team also plans to conduct operator-in-the-loop field studies in plants to quantify detection delays, triage efficiency and human-factor impacts in real-time deployment.
"The system is vital because it is deployment oriented. It integrates with existing machine-learning intrusion detection in substations, microgrids and process plants by inserting a preclassification reactive layer that increases robustness to advanced stealth activity while keeping integration overhead low," said lead author Syed Wali Abbas Rizvi, a Ph.D. student in the Department of Electrical and Computer Engineering and CARES Laboratory researcher.
As cyber-attacks persist, RADIANT offers promise as a new line of defense to help protect the critical infrastructure we rely on.
Also contributing to this research is Ph.D. student Yasir Ali Farrukh.
More information: Irfan Khan et al, RADIANT: Reactive Autoencoder Defense for Industrial Adversarial Network Threats, Computers & Security (2025). DOI: 10.1016/j.cose.2025.104403
Provided by Texas A&M University College of Engineering Citation: A RADIANT future for cybersecurity (2025, October 16) retrieved 16 October 2025 from https://techxplore.com/news/2025-10-radiant-future-cybersecurity.html This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
Explore further
Cyber defense innovation could significantly boost 5G network security
Feedback to editors
