Have you ever wondered how engineers test defenses against massive cyberattacks without accidentally taking down the real networks we rely on every day? As 5G and the emerging 6G networks evolve into the critical, AI-native backbone of our digital infrastructure, they are growing incredibly complex.
Consequently, the risk and financial cost of testing new orchestration policies or security mechanisms directly on live, functioning systems have become prohibitive.
To solve this, researchers are turning to “Network Digital Twins” (NDTs), which are virtual replicas that mirror the behavior of real networks, providing a risk-free sandbox for vital experiments without touching production.
Moving Beyond Rough Sketches
While the concept of a digital twin isn’t entirely new, earlier versions have often acted like rough sketches. Current solutions frequently focus on general, coarse performance predictions rather than accurately replicating the chaotic, high-fidelity reality of dynamic traffic patterns.
A newly published breakthrough is changing that landscape entirely. Researchers have developed an advanced, AI-native Network Digital Twin engine based on a “Graph Transformer” and designed to deeply analyze and reconstruct the intricate, second-by-second spatio-temporal dynamics of 5G traffic.
Instead of just estimating how a network might behave, this AI meticulously learns the underlying statistical structure and natural rhythms of benign, everyday activities -such as streaming a video, and contrasts them with the aggressive, overwhelming data floods typical of a Distributed Denial-of-Service (DDoS) cyberattack.
Unprecedented Accuracy in Simulation
The results of this new approach are genuinely groundbreaking. In rigorous experimental tests using comprehensive telemetry from a real 5G testbed, this AI model achieved an astounding accuracy score of over 98% (with R-squared scores up to 0.9839) in replicating network traffic. It significantly outperformed older, traditional AI baselines, like Long Short-Term Memory (LSTM) networks, proving its unique ability to perfectly mimic the heartbeat of a real network.
When the digital twin simulates an attack, it accurately reproduces the exact data spikes, bottlenecks and signal drops that would occur in reality, preserving the underlying probability distributions and variance of real-world telemetry.
A Risk-Free Future for Cybersecurity
The true novelty of this work lies in this unprecedented level of fidelity. It creates a reusable, secure testing environment capable of generating realistic network traffic traces entirely on demand, without needing access to an actual live network’s measurements.
This innovation represents a massive leap forward for European cybersecurity research, directly supporting the advanced objectives of collaborative efforts like the MARE project.
By providing a genuinely risk-free, highly accurate virtual replica, this digital twin empowers researchers and engineers to safely evaluate robust mitigation strategies, performance and control policies.
Ultimately, this foundational work ensures that as we move into the hyper-connected era of 6G, our critical digital ecosystems will remain remarkably resilient and secure.
Publication Details
Conformal Prediction and Risk-Based Optimization of Service Continuity in 6G Edge Networks
Authors: Athina Vekraki, Maria Christopoulou, Ioannis Vasalos, Michail Alexandros Kourtis, Athanasia Alonisioti, George Xilouris
MARE Partner: National Centre For Scientific Research Demokritos (NCSRD)
Publication:
Abstract: As 5G and emerging 6G networks evolve into highly complex, AI-native infrastructures, the risk and cost of testing new orchestration policies or security mechanisms on live systems become prohibitive. Network Digital Twins (NDTs) have emerged as a critical paradigm for providing risk-free virtual replicas; however, current solutions often focus on coarse performance prediction rather than the high-fidelity replication of dynamic traffic patterns. In this paper, we present an AI-native NDT engine based on Graph Transformers designed to reconstruct the intricate spatio-temporal dynamics of 5G traffic. Using a comprehensive dataset from an Amarisoft-based testbed, we model network telemetry as a k-hop temporal line graph to capture the underlying statistical structure and bitrate oscillations of individual User Equipments (UEs) across both benign streaming and Distributed Denial-of-Service (DDoS) traffic regimes. Experimental results demonstrate that our Graph Transformer architecture achieves high-fidelity reconstruction with R^2 scores up to 0.9839, significantly outperforming traditional Long Short-Term Memory (LSTM) baselines. Furthermore, strong alignment in Cumulative Distribution Function (CDF) analysis confirms the model’s ability to preserve the underlying probability distributions and variance of real-world telemetry. This work provides a foundation for high-fidelity traffic synthesis and reproducible security analysis in future 6G ecosystems.
