Anonymization of Documents for Law Enforcement with Machine Learning

2025-01-13

AI Summary

To address privacy-preserving requirements for document images in law enforcement, this paper proposes a few-shot automated anonymization method: given only a single manually anonymized exemplar, it accurately localizes and adaptively redacts sensitive information (e.g., names, addresses) in semantically similar documents. The method integrates self-supervised visual representation learning (DINOv2), feature-matching-based document instance retrieval, and reference-guided knowledge transfer to construct an end-to-end framework for sensitive region detection and redaction. Evaluated on a custom ground-truth dataset, our approach significantly improves redaction accuracy over fully automatic redaction and template-copying baselines, enhances forensic usability, and achieves several-fold speedup in processing time. This work introduces the first “instance retrieval + reference transfer” few-shot anonymization paradigm, uniquely balancing regulatory compliance, localization precision, and forensic validity.

Technology Category

• Machine Learning: Privacy
• Computer Vision: Image and Video Retrieval
• Natural Language Processing: Safety and Robustness

Application Category

• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

The steadily increasing utilization of data-driven methods and approaches in areas that handle sensitive personal information such as in law enforcement mandates an ever increasing effort in these institutions to comply with data protection guidelines. In this work, we present a system for automatically anonymizing images of scanned documents, reducing manual effort while ensuring data protection compliance. Our method considers the viability of further forensic processing after anonymization by minimizing automatically redacted areas by combining automatic detection of sensitive regions with knowledge from a manually anonymized reference document. Using a self-supervised image model for instance retrieval of the reference document, our approach requires only one anonymized example to efficiently redact all documents of the same type, significantly reducing processing time. We show that our approach outperforms both a purely automatic redaction system and also a naive copy-paste scheme of the reference anonymization to other documents on a hand-crafted dataset of ground truth redactions.

Problem

• Data Privacy
• Automatic Redaction
• Sensitive Information

Innovation

• Machine Learning
• Automatic Redaction
• Pattern Recognition