SAQ2 – Security Architectures and Processes I
Thursday, 4 June 2026, 16:30-18:00, room Sala 6 (1st floor)
Session Chair: Rodrigo Roman Castro (Univ. Malaga, ES)
On Jamming Protection: A Framework for Detection, Identification \& Mitigation
Eleni Chamou and Odysseas Grosomanidis (CERTH, Greece); Alexandros Ioannis Papadopoulos (University of Ioannina, Greece & Information Technologies Institute, Greece); Antonios Lalas (Centre for Research and Technology – Hellas (CERTH), Greece); Konstantinos Votis (Information Technologies Institute, Centre For Research and Technology Hellas, Greece)
Wireless communication systems are increasingly vulnerable to intentional RF jamming, which can severely disrupt data transmission. In this work, we propose STRIX\footnote, an end-to-end framework for jamming detection, identification, and mitigation in MIMO-OFDM networks. STRIX combines eigenvalue-domain source enumeration for robust jamming detection, a ResNet-based direction-of-arrival (DoA) estimator to localize legitimate and malicious transmitters, and a two-stage mitigation strategy using spatial eigenprojection and MVDR beamforming. The framework operates without prior knowledge of the jammer's characteristics and supports constant, periodic, and reactive jamming types, remaining effective under mobility and timing misalignment. Performance evaluations in IEEE 802.11p V2X scenarios demonstrate substantial SINR gains and improved symbol recovery even under strong jamming conditions.
FORTRESS: A Federated Learning and Transformer-Based Zero Trust Framework for O-RAN Security
Harshi Ratnayake (University of Sri Jayewardenepura, Sri Lanka); Yushan Siriwardhana (University of Oulu, Finland); Uditha L. Wijewardhana (University of Sri Jayewardenepura & Faculty of Engineering, Sri Lanka); Madhusanka Liyanage (University College Dublin, Ireland)
Open Radio Access Networks (O-RAN) have emerged as a novel paradigm enabling disaggregated and vendor-interoperable RAN elements. However, increasing open interfaces, multi-vendor interoperability, and programmable interfaces substantially increase the threat surface. This paper proposes FORTRESS, a novel framework that applies the zero trust architecture (ZTA) principles to O-RAN and implements it using federated learning (FL). The ZTA model continuously verifies every O-RAN user, device, and service interaction. FORTRESS automates the security provision by integrating decision-making intelligence into the components of ZTA, and it is trained via FL. Moreover, FL preserves privacy and removes the dependency on centralized data storage and unnecessary traffic transfers. Therefore, this is the first work integrating ZTA components in the O-RAN architecture and utilizing FL to ensure security provision. The paper compares the FL approach against a centralized ML approach where the data needs to be aggregated to a central server. Results from simulation environments demonstrate improved resilience against threats, minimized latency in trust decisions, and efficient policy enforcement.
DDoS Adversarial Attacks Detection with Deep Neural Networks in P4 Programmable Switches
Giulio Zingrillo (ETH Zurich, Switzerland); Layal Ismail and Emilio Paolini (Scuola Superiore Sant'Anna, Italy); Francesco Paolucci and Filippo Cugini (CNIT, Italy); Lorenzo De Marinis (Scuola Superiore Sant'Anna, Italy)
Traditional network intrusion detection systems struggle to keep pace with the rapid evolution of cyber threats. While deep learning based detection systems offer superior performance, their deployment within dataplane is limited by hardware constraints. Moreover, they remain susceptible to adversarial attacks, a novel kind of threat that specifically targets neural networks. This paper proposes an architectural framework for an in-switch, DNN-based NIDS capable of detecting both DDoS and adversarial exploits. The system leverages the LUT distillation technique to map deep neural networks onto P4 programmable switches, such as the Intel Tofino. We design, train and integrate an Adversarial Detector alongside the primary classifier, to make the system resilient to adversarial attacks. Experimental results on the Edge-IIoTset show that the proposed system maintains robust accuracy between 91% and 99% across white-, gray-, and black-box threat models. Hardware validation proves that the framework achieves functional equivalence with reference software while adding only 1-2 μs of latency. This approach provides a scalable and energy-efficient solution for securing high-speed network infrastructure against sophisticated cyber threats.
SEAL: An Open, Auditable, and Fair Data Generation Framework for AI-Native 6G Networks
Sunder Ali Khowaja (Dublin City University, Ireland & University of Sindh, Pakistan); Kapal Dev (Munster Technological University, Ireland); Engin Zeydan (CTTC, Spain); Madhusanka Liyanage (University College Dublin, Ireland)
AI-native 6G networks promise to transform the telecom industry by enabling dynamic resource allocation, predictive maintenance, and ultra-reliable low-latency communications across all layers, which are essential for applications such as smart cities, autonomous vehicles, and immersive XR. However, the deployment of 6G systems results in severe data scarcity, hindering the training of efficient AI models. Synthetic data generation is extensively used to fill this gap; however, it introduces challenges related to dataset bias, auditability, and compliance with regulatory frameworks. In this regard, we propose the Synthetic Data Generation with Ethics Audit Loop (SEAL) framework, which extends baseline modular pipelines with an Ethical and Regulatory Compliance by Design (ERCD) module and a Federated Learning (FL) feedback system. The ERCD integrates fairness, bias detection, and standardized audit trails for regulatory mapping, while the FL enables privacy-preserving calibration using aggregated insights from real testbeds to close the reality-simulation gap. Results show that the SEAL framework outperforms existing methods in terms of Frechet Inception Distance, equalized odds, and accuracy. These results validate the framework's ability to generate auditable and bias-mitigated synthetic data for responsible AI-native 6G development.
Secrecy-Aware Mobility Control for the ORAN Enabled 6G TN-NTN Continuum
Awanthi Jayasundara (University College Dublin, Ireland); Engin Zeydan (CTTC, Spain); Madhusanka Liyanage and Pasika Ranaweera (University College Dublin, Ireland)
Future wireless network deployments are expected to integrate Terrestrial Network Terrestrial Network (TN) and Non-Terrestrial Network Non-Terrestrial Network (NTN) segments to enhance coverage and capacity; however, this convergence also introduces new security challenges for protecting air-interface confidentiality. This study examines the impact of physical-layer security (PLS) in an integrated TN-NTN architecture under realistic mobility and handover dynamics, using an xApp for control. We extend an ns-3 TN-NTN scenario to a network with passive downlink eavesdroppers and introduce a secrecy metric to quantify confidentiality in their presence. Simulation results show that the modelled secrecy-aware xApp improves confidentiality against passive downlink eavesdropping by up to 80%, while preserving service quality continuity. When the xApp decision logic is implemented using an Open Neural Network Exchange (ONNX)-deployed Machine Learning (ML) gating module, the secrecy-outage reduction relative to the non-secrecy baseline reached up to 97.5%. Overall, the results suggest that TN-NTN in a unified continuum should be evaluated not only for coverage extension but also for adaptive security, where the control plane can steer UEs toward serving cells that offer more favorable secrecy conditions under dynamic channel and mobility conditions.
