An Information Theoretic Approach to Operationalize Right to Data Protection

Abhinav Java, Simra Shahid, Chirag Agarwal

Keywords: safety, dataset, unlearnable

TL;DR: A model agnostic framework for generating text data that converges on the training split but fails to generalize on the unseen test dataset!

Abstract:

The widespread practice of indiscriminate data scraping to fine-tune language models raises significant legal and ethical concerns, particularly regarding compliance with data protection laws such as the General Data Protection Regulation (GDPR). This practice often results in the unauthorized use of personal information, prompting growing debate within the academic and regulatory communities.
Recent works have introduced the concept of generating unlearnable datasets (adding imperceptible noise to the clean data), such that the underlying model converges during training but fails to generalize to the unseen test setting. Though somewhat effective, these approaches are predominantly designed for images and are limited by several practical constraints like requiring knowledge of the target model and instability of bi-level optimization. To this end, we introduce RegText, an information-theoretic framework to operationalize the right to data protection in practice. In particular, RegText is a model-agnostic data generation framework that leverages the frequency distribution of tokens within a given dataset to create a ranking system, allowing for the systematic injection of selected words back into the dataset. We demonstrate RegText's utility through rigorous theoretical and empirical analysis of small and large language models. Notably, RegText can restrict newer models like GPT-4o and Llama from learning on our generated data, resulting in a drop in their test accuracy compared to their zero-shot performance and paving the way for generating unlearnable text to protect public data.

Primary Area: alignment, fairness, safety, privacy, and societal considerations

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 9104