DSC 253 Advanced Data-Driven Text Mining Project
Collaborators:
- Jiongli Zhu
- Jintong Luo
- Sihan Wang
- Laurel Li
We investigated the state-of-the-art data poisoning attack based on style attack and style transfer. Our aim is to identify and document the critical limitations that might prevent this (and other similar) attacking strategy from being practical. Through comprehensive empirical analysis and detailed case studies, which are not contained in the original paper, we support our conclusions and provide insights into the effectiveness and detectability of such attacks.