Heterogeneous network intrusion detection via domain adaptation in IoT environment

Abstract Network intrusion detection refers to detect the threaten behaviors in the network to guarantee the network security. Compared with computer network, Internet of Things (IoT) consists of various devices, including computer, smart phone, smart watch, various sensors etc. The data in IoT may be captured from heterogeneous scenes using various devices. The data may follow from different distributions. Most previous works may fail when they are used in heterogeneous scenes of IoT. In order to overcome this issue, this paper designs a heterogeneous network intrusion detection scheme using attention sharing mechanism to implement domain adaptation for the intrusion detection of the data with heterogeneous distributions. The data from heterogeneous IoT devices is projected into the same sharing space via attention sharing to alleviate the bias between the distributions of data from these devices. Thus, the intrusion detection model learnt from the data from a scene can be migrated to another scene. The experiments and simulation demonstrate that the proposed intrusion detection scheme can adapt the changes of IoT scene.