• Spam and Fraud Detection

    We developed new graph-based machine learning methods to detect spams and frauds in various application domains. Read our paper (Distinguished Paper Award Honorable Mention at NDSS'19) for details.

  • Defending against Inference Attacks

    Attacker increasingly misuses machine learning (ML) for automated large-scale inference attacks, e.g., Cambridge Analytica used ML to infer Facebook users' private attributes via their page likes. We proposed the first practical defense against such inference attacks. 

  • Protecting Intellectual Property

    We study how an adversary can compromise the intellectual property of machine/deep learning models and how to protect the intellectual property. Read our paper on stealing hyperparameters in machine/deep learning models.  

     

  • Securing Deep Neural Networks against Adversarial Examples

    We proposed a randomized defense against adversarial examples without sacrificing accuracy of normal examples. 

Welcome to Neil Gong Research Group

Welcome to Neil Gong's research group in the Department of Electrical and Computer Engineering at Duke University. Our research group broadly studies cybersecurity and privacy with a recent focus on the intersections between cybersecurity, privacy, and artificial intelligence. Our group is also interested in mobile and IoT security, privacy, and forensics. On one hand, we leverage various artificial intelligence techniques, including but not limited to, machine learning, deep learning, probabilistic graphical models, network science, optimization, and natural language processing, to study cybersecurity and privacy. On the other hand, we build secure and privacy-preserving artificial intelligence techniques via robust optimization, differential privacy, statistics, trusted hardware, etc..