Ada: Adversarial learning based data augmentation for malicious users detection

Jia Wang, Min Gao, Zongwei Wang, Chenghua Lin, Wei Zhou, Junhao Wen

Abstract: Highlights•A detection model dealing with the lack of labeled malicious users is proposed.•DCGAN based data augmentation simulates the distribution of real users.•Reliable user embeddings are based on potential relationships and interactive structure.•Experiments show the superiority of the data augmentation and user embeddings.