My research interests include AI security, data privacy, Web3 and blockchain systems, with the focus on AI security and model robustness, data privacy in AI and various fields, Web3 and decentralized identity (DID), blockchain security and system development.

Towards Adversarial Robustness with Fine-grained Unrestricted Adversarial Examples

We introduce a new generative network with separated latent features to construct attacks and defenses for a “smile” detector.

From Common to Individual: Geometry-Inspired Perturbations against Deep Neural Networks

We propose a new adversarial attack method called GeoFool, GeoFool can generate Common knowledge and individual knowledge.

Generating Unrestricted Adversarial 3D Point Cloud with Graph-Convolutional GAN

We propose a new generative adversarial attack algorithm with graph-convolutional condition GAN, which aims towards realistic and strong adversarial attack against 3D point cloud deep learning model.

Abnormal Traffic Detection: Traffic Feature Extraction and DAE-GAN with Efficient Data Augmentation

We introduce a DAE-GAN model and a data-augmentation-based training algorithm for detecting abnormal traffic.