Robust ML Auditing using Prior Knowledge

2025-05-07

AI Summary

Machine learning decision systems may be maliciously manipulated during fairness auditing—i.e., platforms selectively alter outputs only for audit queries to pass scrutiny while retaining discriminatory behavior toward real users. Method: We propose the first regulator-prior-driven, manipulation-resistant fairness auditing framework. We formally define “prior-driven non-manipulable auditing conditions,” expose the vulnerability of public priors to strategic deception, and establish theoretical bounds for robust auditing under true data-distribution priors. Using game-theoretic modeling, probabilistic inference, and formal verification, we derive the maximum hidden unfairness degree that platforms can conceal. Results: Empirical evaluation on Adult and COMPAS datasets demonstrates significantly improved detection of covert discrimination. Our framework achieves the first provably secure fairness audit—guaranteeing audit integrity against strategic platform manipulation while preserving statistical validity.

Technology Category

• Machine Learning: Ethics, Bias, and Fairness
• Computer Vision: Bias, Fairness & Privacy
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

The rapid adoption of ML decision-making systems across products and services has led to a set of regulations on how such systems should behave and be built. Among all the technical challenges to enforcing these regulations, one crucial, yet under-explored problem is the risk of manipulation while these systems are being audited for fairness. This manipulation occurs when a platform deliberately alters its answers to a regulator to pass an audit without modifying its answers to other users. In this paper, we introduce a novel approach to manipulation-proof auditing by taking into account the auditor's prior knowledge of the task solved by the platform. We first demonstrate that regulators must not rely on public priors (e.g. a public dataset), as platforms could easily fool the auditor in such cases. We then formally establish the conditions under which an auditor can prevent audit manipulations using prior knowledge about the ground truth. Finally, our experiments with two standard datasets exemplify the maximum level of unfairness a platform can hide before being detected as malicious. Our formalization and generalization of manipulation-proof auditing with a prior opens up new research directions for more robust fairness audits.

Problem

• Addressing manipulation risks in ML fairness audits
• Preventing platforms from fooling auditors using prior knowledge
• Establishing conditions for robust manipulation-proof auditing

Innovation

• Uses prior knowledge for manipulation-proof auditing
• Avoids reliance on public datasets to prevent fooling
• Formally establishes conditions for preventing audit manipulations