Self-Adaptive Cybersecurity Intelligence Using Graph Neural Networks and AI Agents for Critical Infrastructure Protection

Authors

  • Chaitanya Nagabhyru 12-1006, Rahmath Nagar 2nd Line, Chilakaluripet, Guntur Pin: 522616, Andhra Pradesh, India

DOI:

https://doi.org/10.66838/sauc.6546

Keywords:

Graph Neural Networks, Cybersecurity, Critical Infrastructure, Novelty Detection, Explainable Alerts, Adversarial Robustness, Human-in-the-Loop Response, AI Agents.

Abstract

 Hospitals, water systems, electric utilities, transit agencies and public-sector networks are attacked through chains of ordinary-looking relationships: a stolen credential logs into a workstation, the workstation reaches a server, the server talks to a controller. Detectors that score each device or log line in isolation miss these chains, and automated responses that act without oversight can disrupt the very services they protect. This paper presents a defensive framework that represents users, hosts, operational-technology (OT) devices and external endpoints as a heterogeneous time-windowed graph, classifies known attack behavior with a relational graph neural network, flags previously unseen behavior with a reconstruction-based novelty module, explains each alert by the relationships that drive it, and recommends risk-tiered response actions that require human approval when they could affect critical assets. ...

Downloads

Download data is not yet available.

Global Statistics ℹ️

Cumulative totals since publication
12
Views
7
Downloads
19
Total

Downloads

Published

2025-01-30

How to Cite

Chaitanya Nagabhyru. (2025). Self-Adaptive Cybersecurity Intelligence Using Graph Neural Networks and AI Agents for Critical Infrastructure Protection. Street Art & Urban Creativity, 11(1), 252–268. https://doi.org/10.66838/sauc.6546

Issue

Section

Research articles