Heterogeneous data processing and network attack detection based on two-level and multi-segment model

Because of the heterogeneous character of network data, a reasonable method is required to integrate the heterogeneous data in network attack detection. In this paper, we propose a two-level and multi-segment model for network attack detection. Meanwhile, the parallel training method is also involved. In the experiments, DARAP 1998 dataset has been used which is a public cyber attacks dataset. The experimental result shows that when facing the heterogeneous data, the two-level and multi-segments model has an effective performance that the proposed model obtains a better detecting accuracy and recall than the fully connected neural network.