Intrusion detection using quantum generative adversarial networks: a federated approach with noisy simulators

In today's increasingly complex digital environments, Intrusion Detection Systems (IDS) play a crucial role in ensuring network security. Traditional machine learning approaches struggle with challenges such as handling high-dimensional data and maintaining performance on imbalanced datasets. Generative Adversarial Networks (GANs) offer a viable alternative by enhancing data generation, but their conventional implementations are computationally intensive and strive with capturing intricate data distributions. Quantum GANs (QGANs) leverage quantum computing to address these limitations, while a distributed approach enhances load balancing. Tested on the NSL-KDD dataset, the proposed model effectively learns the distribution of benign data using a federated hybrid QGAN architecture, which integrates quantum generators with classical discriminators. Additionally, the model has been evaluated on a quantum noisy simulator to assess performance variations under noise conditions.