Code supporting the publication: CatBack: Universal Backdoor Attacks on Tabular Data via Categorical Encoding

In this repo we have included the code for the paper Tajalli et al. “CatBack: Universal Backdoor Attacks on Tabular Data via Categorical Encoding.” Network and Distributed System Security (NDSS) Symposium 2026, Internet Society, 2026, doi:10.14722/ndss.2026.231469. This paper introduces CatBack, a novel backdoor attack specifically designed for tabular data containing both numerical and categorical features. The main research theme addresses a significant gap in backdoor attack research, which has predominantly focused on homogeneous data like images, leaving tabular data vulnerabilities largely unexplored.