I am currently a second-year graduate student at the Xiamen University Institute of Artificial Intelligence, focusing on research in neural network robustness, particularly adversarial purification.

I earned my Bachelor of Engineering degree in Data Science and Big Data Technology from the School of Computer Science and Big Data at Fuzhou University. I am now pursuing my masterโ€™s degree at the Intelligent Multimedia Laboratory at Xiamen University under the guidance of Associate Professor Zhiming Luo.

๐Ÿ”ฅ News

  • Nothing has happened recently.๐Ÿซก

๐Ÿ“ Publications

MANI-Pure: Magnitude-Adaptive Noise Injection for Adversarial Purification

Xiaoyi Huang, Junwei Wu, Kejia Zhang, Carl Yang, Zhiming Luo

sym
Contributions
  • We further reveal distributional differences between adversarial and clean samples in the magnitude spectrum.
  • The proposed MANI-Pure framework combines magnitude-adaptive diffusion with frequency-domain purification, achieving a principled balance between semantic fidelity and perturbation mitigation.
  • Extensive experiments across datasets, attacks, and backbones demonstrate the superiority of our method.

๐Ÿ“– Educations

  • 2024.09 - now, Xiamen University, Institute of Artificial Intelligence, Artificial Intelligence, Intelligent Multimedia Laboratory.
  • 2020.09 - 2024.06, Fuzhou University, School of Computer Science and Big Data, Data Science and Big Data Technology, Bachelor of Engineering.

๐ŸŽ– Honors and Awards

  • 2023.05 Meritorious Winner, Mathematical Contest in Modeling (MCM/ICM)
  • 2023.07 Honorable Mention, Mathematical Contest in Modeling (MCM/ICM)