Research

Currently, I’m working on the following projects. Feel free to drop by or send me emails if you’re interested in them.

• Detecting poisoning attacks on ML models. We mainly focus on the statistic difference of poisoning samples/models.
• Poisoning attacks on contrastive learning, CLIP, etc.
• Security and privacy of federated learning (FL). Our interests in FL are that it is distributed and vulnerable to insider attacks. At the same time, FL is subject to performance and communication constrains. These make it challenging and important to work around both security and system design goals.
• Jailbreak and prompt injection attacks on LLMs.
• Federated IoT systems. We want to test drive FL on IoT systems like smart home or intrusion detection systems.
• Coining AI models such as contrastive learning, transfer learning, multitasking learning for security applications such as user authentication, malware detection.
• Practical differential privacy (DP/LDP) systems on real data such as mobile sensor data, indoor localization data, medical data.
• Coining privacy-aware NLP models for medical EHR.
• GNN-based causal inference for data stream applications such as prediction and recommendation.

Members

I’m very lucky to work with the following talented students. Join us if you have interests in our projects!

• PhD students
  • Shixiong Li (2022’fall to present, BS/MS from Southwest Jiaotong Univ.)
  • Xingyu Lyu (2022’fall to present, BS from Shanghai Univ., MS from Guangzhou Univ.)
• Undergrad students at UML
  • Sean Cox (2024’spring, malware generation via LLMs, UML HONOR Thesis)
  • Gurpreet Singh (2023’summer, Contrastive learning, UML Immersive Scholar)
  • Michelle Ly (2022’fall to 2023’spring, NLP, UML Immersive Scholar)
  • Jared Q Widberg (2022’spring to 2022’fall, Reverse engineering, UML HONOR Thesis)
• Alumni
  • Sean Cox (Spring 2024 BS. First stop: NETSCOUT)
  • Manita (Spring 2023 MS. First stop: SANBlaze Technology)
  • Jared Q Widberg (Spring 2023 BS. First stop: BAE Systems, Inc.)
• Exchange students/scholars
  • Tsz Him Shek (2023’summer, Contrastive learning, The Chinese University of Hong Kong; Next stop: MSc at Applied Computing, U Toronto)

Collaborations

I work closely with Dr. Tao Li at CIT Purdue, Dr. Danjue Chen at NCSU, Dr. Ning Wang at USF, and Dr. Sashank Narain at UML.