Neil Zhenqiang Gong

Investigating vulnerabilities in audio watermarking methods against real-world threats.

MLLMs face new threats as safe prompts can be falsely refused.

A framework enabling data owners to check unauthorized use of their data in ML.