Attack detection method and device for heterogeneous network data

The invention relates to an attack detection method and device oriented to heterogeneous network data. The method comprises the following steps: acquiring a pre-trained network flow classification model, a real-time source network flow data sample and a real-time target network flow data sample; the network flow classification model comprises a source network flow classification model and a target network flow classification model based on an online learning model; respectively performing online prediction on the source network flow data sample and the target network flow data sample through a network flow classification model to obtain a source prediction result of the source network flow data sample and a target prediction result of the target network flow data sample; and according to the source prediction result and the target prediction result, respectively carrying out concept drift detection on the source network flow data sample and the target network flow data sample, and updating the network flow cl