AAE-DSVDD: A one-class classification model for VPN traffic identification

Sicai Lv, Chao Wang, Zibo Wang, Shuo Wang, Bailing Wang, Yongzheng Zhang

Abstract: Highlights•We introduce AAE-DSVDD for VPN traffic identificaiton and compared it with one-class models and supervised learning.•We optimized the output of DSVDD by distributional constraints to achieve a better performance, and it alleviate the hypersphere collapse problem.•We confirmed that automated feature extraction via neural networks is more effective than using the original ISCXVPN provided features.