Variational Few-Shot Learning for Microservice-Oriented Intrusion Detection in Distributed Industrial IoT

Along with the popularity of Internet Things (IoT) techniques several computational paradigms, such as cloud and edge computing, microservice has been viewed a promising architecture in large-scale application design deployment. Due to limited computing ability devices distributed IoT, only small scale data can be used for model training. In addition, most machine-learning-based intrusion detection methods are insufficient when dealing imbalanced dataset under resources. this article, we propose an optimized intra/inter-class-structure-based variational few-shot learning (OICS-VFSL) overcome specific out-of-distribution problem learning, improve microservice-oriented IoT systems. Following newly designed VFSL framework, intra/inter-class optimization scheme is developed using reconstructed feature embeddings, which intra-class distance based on approximation during variation Bayesian process, while inter-class maximization similarities concatenation process. An intelligent algorithm is, then, introduced multiclass classification via nonlinear neural network. Evaluation experiments conducted two public datasets demonstrate effectiveness our proposed model, especially detecting novel attacks extremely data, compared four baseline methods.