Mengxin Zheng

FHE-Coder: Benchmarking Secure Agentic Code Generation for Fully Homomorphic Encryption

PRO: Enabling Precise and Robust Text Watermark for Open-Source LLMs

RPP: A Certified Poisoned-Sample Detection Framework for Backdoor Attacks under Dataset Imbalance

TwinShield: Secure Foundation Model Execution by Unifying TEEs and Crypto-protected Accelerators

DictPFL: Efficient and Private Federated Learning on Encrypted Gradients

CipherPrune: Efficient and Scalable Private Transformer Inference

DictPFL: Efficient and Private Federated Learning on Encrypted Gradients

BadRAG: Identifying Vulnerabilities in Retrieval Augmented Generation of Large Language Models