Reproduce “Towards Generating Adversarial Examples on Mixed-type Data” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Han Xu, Menghai Pan, Zhimeng Jiang, Huiyuan Chen, Xiaoting Li, Mahashweta Das, Hao Yang

M-Attack generates adversarial examples for classifiers on mixed numerical and categorical records. It jointly updates numerical values and categorical-choice distributions, while penalizing a mixed-type distance intended to preserve clean-data relationships. Across three datasets, it reports higher success than sequential baselines on CIS Fraud and Criteo, while all methods do well on Home Credit; detection evidence is limited to KDE-based measures.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.