A Few-Shot Intrusion Detection Model for the Internet of Things

The development of Internet technology has enabled the world to enter the era of "interconnectedness of everything", greatly improving the convenience of production and life. However, the challenge of network security threats hidden inside IoT is not optimistic, and network attacks are often low-frequency, sudden and difficult to recognize. For this reason, this paper proposes a few-shot intrusion detection model for IoT, which consists of three sub-modules, namely, data augmentation, type conversion, and image categorization, to improve the detection accuracy while ensuring the detection rate. Experiments on two datasets, CICIDS2018 and N-BaLoT, demonstrate a certain degree of improvement over the baseline model in terms of accuracy, precision, recall, and F1-Score. It is not only applicable to real IoT environments, but also generalized to various network types.