Deep generative models as an adversarial attack strategy for tabular machine learning

TL;DR

This paper explores how deep generative models can be adapted as adversarial attack strategies for tabular machine learning, addressing unique challenges in generating realistic, domain-constrained adversarial examples.

Contribution

It introduces four adapted tabular deep generative models as adversarial methods and evaluates their effectiveness in creating realistic, domain-compliant adversarial examples.

Findings

AdvDGMs successfully generate realistic adversarial examples for tabular data.

The methods maintain domain constraints in adversarial samples.

Evaluation shows improved attack effectiveness over baseline methods.

Abstract

Deep Generative Models (DGMs) have found application in computer vision for generating adversarial examples to test the robustness of machine learning (ML) systems. Extending these adversarial techniques to tabular ML presents unique challenges due to the distinct nature of tabular data and the necessity to preserve domain constraints in adversarial examples. In this paper, we adapt four popular tabular DGMs into adversarial DGMs (AdvDGMs) and evaluate their effectiveness in generating realistic adversarial examples that conform to domain constraints.