My Projects

LLM for Network Traffic in Cybersecurity

• The challenge lies in heterogeneous traffic data including both plaintext and ciphertext.
• Developed a foundation model that leverages the T5 architecture to learn pre-trained representations from massive traffic data.
• Apply the proposed models to 20 downstream tasks, including IoT attack detection and Malware detection.

Physcis-based Learning for Cyber-Physical Systems

• Incorporated the laws of physics into deep learning models to improve their interpretability and generalization
• Developed Taylor Neural Networks without using activation functions
• Apply the proposed models to cyber-physical systems, AD, and climate changes

Concept-based Interpretable Machine Learning

• Proposed a novel ControlVAE that combines control theory with Variational Autoencoders to disentangle the latent factors.
• Adopted disentangled latent factors to explain the prediction results

Privacy-Preserving Federated Learning

• Developed data-free one-shot federated learning for heterogeneous data
• Developed adaptive gradient protection method to achieve a good trade-off between interpretability and privacy protection