Quantum GANs for Smarter Network Anomaly Detection
As modern communication networks grow in complexity, traditional anomaly detection techniques face challenges in identifying sophisticated threats such as distributed denial-of-service (DDoS) or stealth attacks. In our latest study, we explore how quantum machine learning can offer a new frontier for cybersecurity analytics.
In our recent paper, “Enhancing Network Anomaly Detection with Quantum GANs and Successive Data Injection for Multivariate Time Series” [arXiv], Hammami, Cherkaoui, and Wang present a Quantum Generative Adversarial Network (QGAN) designed for detecting anomalies in multivariate network time-series data. The model integrates variational quantum circuits, data re-uploading, and a novel Successive Data Injection (SuDaI) technique that efficiently maps high-dimensional classical data into quantum states—overcoming current hardware limits such as the small number of available qubits.
Figure 1 — Quantum Discriminator Circuit.
The discriminator circuit used in the proposed Quantum GAN architecture for network anomaly detection. It evaluates generated samples against real multivariate time-series data to guide the generator toward realistic outputs.
Trained on the CIC-CIDDS-2018 dataset, the proposed model achieves 99% accuracy and an F1-score of 0.98, matching the performance of deep learning models that use thousands of parameters while requiring only 80 trainable parameters. Even under simulated quantum noise, the QGAN maintained robust performance, highlighting its potential for real-world deployment on near-term quantum devices.
This research demonstrates that quantum-enhanced learning can deliver compact, efficient, and resilient models for network security in 6G and beyond, opening the door to hybrid quantum–classical cybersecurity systems.
Authors: Wajdi Hammami, Soumaya Cherkaoui, and Shengrui Wang (Polytechnique Montréal, Université de Sherbrooke).