SemantiGuard: AI-Driven Cyber Threat Intelligence for 6G Semantic Communications

Mikhail, Evram Magdy, Shah, Syed T., Malik, Hassan ORCID: https://orcid.org/0000-0002-8564-3683, Saad, Walid M. and Shawky, Mahmoud A. (2026) SemantiGuard: AI-Driven Cyber Threat Intelligence for 6G Semantic Communications. In: IEEE Vehicular Technology Conference (Fall) 2026. UNSPECIFIED. (In Press)

Preview

PDF (VTS_Conference_2026__Revised_Final_Jun26) - Accepted Version Available under License Creative Commons Attribution. Download (3MB) | Preview

Abstract

As 6G networks shift from bit-oriented communication toward Semantic Communication (SemCom), the physical layer becomes vulnerable to a new type of threat called Semantic Adversarial Attacks. In these attacks, adversaries generate precise, low-energy perturbations to alter the meaning of transmitted data while evading detection by traditional energy based security systems. This paper presents SemantiGuard, a cyber threat intelligence (CTI) framework that uses deep joint source-channel coding (Deep JSCC) with unsupervised geometric anomaly detection mechanism to secure the SemCom systems by monitoring the high-dimensional latent manifold, our framework is able to detect the legitimate signals as well as the malicious gradient-optimized perturbations. Using a developed simulation environment, we demonstrate that SemantiGuard achieves more than 84% detection accuracy even in high noise radio environments. Our findings highlights the importance of spatial-aware security layers to ensure the integrity of the critical control messages in the future of autonomous and energy sustainable 6G network architectures.

Item Type:	Book Section
Faculty \ School:	Faculty of Science > School of Computing Sciences
UEA Research Groups:	Faculty of Science > Research Groups > Cyber Intelligence and Networks Faculty of Science > Research Groups > Data Science and AI
Depositing User:	LivePure Connector
Date Deposited:	22 Sep 2026 08:56
Last Modified:	22 Sep 2026 08:56
URI:	https://ueaeprints.uea.ac.uk/id/eprint/104604
DOI:

Downloads

Actions (login required)

View Item