mlsanityData to 7 Oct 2026

Robust ML Auditing using Prior Knowledge

arXiv 2505.04796 ↗First posted 7 May 2025cs.LGcited by 6

Not on the map. Why?

Abstract (arXiv)

Among the many technical challenges to enforcing AI regulations, one crucial yet underexplored problem is the risk of audit manipulation. 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 illustrate 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.

Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Builds on 40 of its 64 references that are in the map. First posted on arXiv 7 May 2025.

Builds on

40 of 64 references.

References · 40

  1. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  2. 2016Equality of Opportunity in Supervised Learningcited by 2,000
  3. 2019A Survey on Bias and Fairness in Machine Learningcited by 1,911
  4. 2021Retiring Adult: New Datasets for Fair Machine Learningcited by 443
  5. 2019Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methodscited by 440
  6. 2020Fairness in Machine Learning: A Surveycited by 369
  7. 2021AI and the Everything in the Whole Wide World Benchmarkcited by 340
  8. 2018Multiaccuracy: Black-Box Post-Processing for Fairness in Classificationcited by 221
  9. 2019Inherent Tradeoffs in Learning Fair Representationscited by 144
  10. 2019Wasserstein Fair Classificationcited by 131
  11. 2019Fairwashing: the risk of rationalizationcited by 81
  12. 2020Fairwashing Explanations with Off-Manifold Detergentcited by 71
28 more
  1. 2022Outsider Oversight: Designing a Third Party Audit Ecosystem for AI Governancecited by 67
  2. 2017Distill-and-Compare: Auditing Black-Box Models Using Transparent Model Distillationcited by 62
  3. 2022Leveling Down in Computer Vision: Pareto Inefficiencies in Fair Deep Classifierscited by 53
  4. 2023Who Audits the Auditors? Recommendations from a field scan of the algorithmic auditing ecosystemcited by 49
  5. 2024AI auditing: The Broken Bus on the Road to AI Accountabilitycited by 42
  6. 2022Understanding Practices, Challenges, and Opportunities for User-Engaged Algorithm Auditing in Industry Practicecited by 30
  7. 2022Active Fairness Auditingcited by 27
  8. 2020Justicia: A Stochastic SAT Approach to Formally Verify Fairnesscited by 25
  9. 2020A Statistical Test for Probabilistic Fairnesscited by 25
  10. 2023Auditing Fairness by Bettingcited by 24
  11. 2021Testing Group Fairness via Optimal Transport Projectionscited by 23
  12. 2021Characterizing the risk of fairwashingcited by 21
  13. 2022Attribution-based Explanations that Provide Recourse Cannot be Robustcited by 20
  14. 2021Algorithmic Fairness Verification with Graphical Modelscited by 16
  15. 2024From Transparency to Accountability and Back: A Discussion of Access and Evidence in AI Auditingcited by 15
  16. 2019Faking Fairness via Stealthily Biased Samplingcited by 14
  17. 2024Trustless Audits without Revealing Data or Modelscited by 14
  18. 2024FairProof : Confidential and Certifiable Fairness for Neural Networkscited by 13
  19. 2022Optimal Transport of Classifiers to Fairnesscited by 12
  20. 2023Sociotechnical Audits: Broadening the Algorithm Auditing Lens to Investigate Targeted Advertisingcited by 12
  21. 2021Fairness seen as Global Sensitivity Analysiscited by 8
  22. 2021Responsible and Regulatory Conform Machine Learning for Medicine: A Survey of Challenges and Solutionscited by 7
  23. 2024Active Fourier Auditor for Estimating Distributional Properties of ML Modelscited by 6
  24. 2022XAudit : A Theoretical Look at Auditing with Explanationscited by 5
  25. 2024Under manipulations, are some AI models harder to audit?cited by 5
  26. 2024Reasons to Doubt the Impact of AI Risk Evaluationscited by 4
  27. 2023Leveraging Imperfect Sources to Detect Fairwashing in Black-Box Auditingcited by 2
  28. 2024Fairness Auditing with Multi-Agent Collaborationcited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 4.

6 papers.

  1. 2025Auditing Pay-Per-Token in Large Language Modelscited by 4
  2. 2025Exposing the Illusion of Fairness: Auditing Vulnerabilities to Distributional Manipulation Attackscited by 2
  3. 2026Manipulation-Proof Oblivious Audits against Deceptive Model Providerscited by 1
  4. 2026The Evaluation Game: Beyond Static LLM Benchmarkingcited by 0
  5. 2026Fairness Auditing: Lower Bounds on Company Manipulationcited by 0
  6. 2026Strategic Self-Consistencycited by 0