Publications
A collection of my research work.
BackdoorDM: A Comprehensive Benchmark for Backdoor Learning in Diffusion Model
Weilin Lin*, Nanjun Zhou*, Yanyun Wang, Jianze Li, Hui Xiong, Li Liu†
NeurIPS 2025
BackdoorDM is the first comprehensive benchmark for backdoor learning on diffusion models, integrating SOTA attacks, defenses, unified evaluation metrics, and multimodal analysis tools to enable fair and systematic comparison.
Gradient norm-based fine-tuning for backdoor defense in automatic speech recognition
Nanjun Zhou*, Weilin Lin*, Li Liu†
ICASSP 2025
GN-FT is a gradient-norm–based fine-tuning defense that selectively suppresses backdoored neurons, providing the first effective and specialized protection against audio backdoor attacks.