We are glad to inform you that our project partners, K3Y and UBITECH, in collaboration with Kingston University London, Democritus University of Thrace, and The University of Western Macedonia have published the paper “Trustworthy Analytics in ETSI ZSM: A 5G Security Case Study” in the IEEE Open Journal of the Communications Society.
The paper investigates the role of cross-domain storage, communication, and analytics services in reinforcing security in the 5G core. Specifically, the authors discuss the deployment of a User Plane Function (UPF) equipped with an AI-driven intrusion detection system (IDS), to ensure security prior to onboarding. This research introduces a novel security game that models the strategies of attackers and defenders, wherein attackers use adversarial techniques to mislead defenders, while defenders use early detection and adversarial training to protect the AI models. In this security game, the Bayesian Nash equilibrium derives the optimal strategies for both parties. Finally, experimental findings highlight how cross-domain analytics can effectively evaluate AI security risks and improve network resilience. While the attacker is able to degrade AI intrusion detection performance below that of a random classifier, essentially disarming the defender, we show that adversarial training not only protects the defender but further improves IDS performance by 3%.
The publication (DOI: 10.1109/OJCOMS.2024.3505555) is available online, and you may access it by following this link: https://ieeexplore.ieee.org/document/10766438
Congratulations to the authors!
