Synthetic Tabular Data Detection In the Wild

2025-03-03

AI Summary

This paper addresses the robust detection of synthetic samples in heterogeneous-structure tabular data. We propose four structure-agnostic detectors and, for the first time, systematically validate cross-structural transferability for synthetic table detection. Our method employs lightweight preprocessing and six progressive “wildness” evaluation protocols to reliably identify synthetic tables with unseen structures under realistic, dynamic scenarios. Key contributions are threefold: (1) We identify fundamental challenges in cross-table transfer, demonstrating that simple preprocessing enables only limited generalization; (2) We empirically verify that structure-agnostic modeling breaks structural dependency, establishing a novel paradigm for detecting wild tabular data; (3) We introduce the first benchmark and evaluation framework specifically designed for synthetic table detection under structural heterogeneity, significantly improving detection robustness and generalization across diverse table schemas.

Technology Category

• Natural Language Processing: Safety and Robustness
• Machine Learning: Structured Learning
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Web Mining and Content Analysis: Bridging structured and unstructured data
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Detecting synthetic tabular data is essential to prevent the distribution of false or manipulated datasets that could compromise data-driven decision-making. This study explores whether synthetic tabular data can be reliably identified across different tables. This challenge is unique to tabular data, where structures (such as number of columns, data types, and formats) can vary widely from one table to another. We propose four table-agnostic detectors combined with simple preprocessing schemes that we evaluate on six evaluation protocols, with different levels of ''wildness''. Our results show that cross-table learning on a restricted set of tables is possible even with naive preprocessing schemes. They confirm however that cross-table transfer (i.e. deployment on a table that has not been seen before) is challenging. This suggests that sophisticated encoding schemes are required to handle this problem.

Problem

• Detecting synthetic tabular data to prevent false datasets.
• Exploring reliable identification across diverse table structures.
• Challenges in cross-table transfer for unseen data.

Innovation

• Proposes four table-agnostic synthetic data detectors
• Evaluates detectors using six wildness protocols
• Highlights need for sophisticated encoding schemes