Federated Multi-Discriminator BiWGAN-GP Based Collaborative Anomaly Detection for Virtualized Network Slicing

Virtualized network slicing allows a multitude of logical networks to be created on common substrate infrastructure support diverse services. A virtualized slice is combination multiple virtual functions, which run machines (VMs) as software applications by virtualization techniques. As the performance slices hinges normal running VMs, detecting and analyzing anomalies in VMs are critical. Based three-tier management framework slicing, we first develop federated learning (FL) based distributed VM anomaly detection framework, enables managers collaboratively train global model while keeping metrics data locally. The high-dimensional, imbalanced, features scenarios invalidate existing models. Considering powerful ability generative adversarial (GAN) capturing distribution from complex data, design new multi-discriminator Bidirectional Wasserstein GAN with Gradient Penalty (BiWGAN-GP) learn high-dimensional resource datasets that spread monitors. BiWGAN-GP can trained over sources, avoids high communication computation overhead caused centralized collection processing local data. We define an score discriminant criterion quantify deviation learned detect abnormal behaviors arising VMs. efficiency effectiveness proposed collaborative algorithm validated through extensive experimental evaluation real-world dataset.