Denoising Diffusion Generative Adversarial Network Integrating Multi-Scale CNN
Yanmin Cao, Fenglian Dong, Zhiyi Xin, Zhiwei Wei, Longkun Han · 2024
With the progress of networking technology, network attack become increasingly diverse and lead to more severe consequences from emerging attacks. The unbalanced data has a great impact on modeling accuracy. In the vast amount of network traffic, the number of different network attack samples varies greatly. Therefore, balancing the network traffic dataset and detecting different network attacks is an urgent problem to be addressed. To tackle this challenge, this paper proposes a denoising diffusion generative adversarial network (DDGAN) integrating multi-scale CNN (MCNN) (DDGAN-MCNN). The DDGAN augments the minority class samples in the traffic dataset and generate synthetic samples to achieve data balancing and enhance the detection rate of minority class intrusions in detection process. Furthermore, a MCNN model is employed as the classifier to identify different intrusions, presenting a comprehensive DDGAN-MCNN intrusion detection method. Experimental results on the UNSW-NB15 and CIC-IDS2017 network traffic datasets demonstrate our method remain higher accuracy compared to other data generation and detection models, that validate the superiority and feasibility of the proposed approach.