TAROT: Task-Oriented Authorship Obfuscation Using Policy Optimization Methods (2407.21630v1)
Abstract: Authorship obfuscation aims to disguise the identity of an author within a text by altering the writing style, vocabulary, syntax, and other linguistic features associated with the text author. This alteration needs to balance privacy and utility. While strong obfuscation techniques can effectively hide the author's identity, they often degrade the quality and usefulness of the text for its intended purpose. Conversely, maintaining high utility tends to provide insufficient privacy, making it easier for an adversary to de-anonymize the author. Thus, achieving an optimal trade-off between these two conflicting objectives is crucial. In this paper, we propose TAROT: Task-Oriented Authorship Obfuscation Using Policy Optimization, a new unsupervised authorship obfuscation method whose goal is to optimize the privacy-utility trade-off by regenerating the entire text considering its downstream utility. Our approach leverages policy optimization as a fine-tuning paradigm over small LLMs in order to rewrite texts by preserving author identity and downstream task utility. We show that our approach largely reduce the accuracy of attackers while preserving utility. We make our code and models publicly available.
- A multifaceted framework to evaluate evasion, content preservation, and misattribution in authorship obfuscation techniques. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 2391–2406, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.
- Adversarial stylometry: Circumventing authorship recognition to preserve privacy and anonymity. ACM Trans. Inf. Syst. Secur., 15(3).
- European Parliament and Council of the European Union. 2016. General data protection regulation (GDPR).
- BertAA : BERT fine-tuning for authorship attribution. In Proceedings of the 17th International Conference on Natural Language Processing (ICON), pages 127–137, Indian Institute of Technology Patna, Patna, India. NLP Association of India (NLPAI).
- Jamdec: Unsupervised authorship obfuscation using constrained decoding over small language models. arXiv preprint arXiv:2402.08761.
- Deberta: Decoding-enhanced bert with disentangled attention. In International Conference on Learning Representations.
- Author masking through translation. In Working Notes of CLEF 2016 - Conference and Labs of the Evaluation forum, Évora, Portugal, 5-8 September, 2016, volume 1609 of CEUR Workshop Proceedings, pages 890–894. CEUR-WS.org.
- Paraphrasing evades detectors of ai-generated text, but retrieval is an effective defense. arXiv preprint arXiv:2303.13408.
- Keep it simple: Unsupervised simplification of multi-paragraph text. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 6365–6378, Online. Association for Computational Linguistics.
- Towards general text embeddings with multi-stage contrastive learning.
- Anonymisation models for text data: State of the art, challenges and future directions. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 4188–4203, Online. Association for Computational Linguistics.
- Authorship style transfer with policy optimization.
- A girl has no name: Automated authorship obfuscation using mutant-x. Proceedings on Privacy Enhancing Technologies, 2019(4):54–71.
- Deep reinforcement learning-based text anonymization against private-attribute inference. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2360–2369, Hong Kong, China. Association for Computational Linguistics.
- Frederick Mosteller and David L. Wallace. 1963. Inference in an authorship problem. Journal of the American Statistical Association, 58(302):275–309.
- Training language models to follow instructions with human feedback.
- GloVe: Global vectors for word representation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1532–1543, Doha, Qatar. Association for Computational Linguistics.
- Author obfuscation: Attacking the state of the art in authorship verification. In Conference and Labs of the Evaluation Forum.
- Direct preference optimization: Your language model is secretly a reward model.
- Learning universal authorship representations. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 913–919, Online and Punta Cana, Dominican Republic. Association for Computational Linguistics.
- Effects of age and gender on blogging. In AAAI Spring Symposium: Computational Approaches to Analyzing Weblogs.
- Proximal policy optimization algorithms.
- Authorship attribution with topic models. Computational Linguistics, 40(2):269–310.
- The GDPR and unstructured data: is anonymization possible? International Data Privacy Law, 12(3):184–206.
- Alison: Fast and effective stylometric authorship obfuscation. In AAAI.
- Adversarial authorship attribution for deobfuscation. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7372–7384, Dublin, Ireland. Association for Computational Linguistics.
- Character-level convolutional networks for text classification. In Advances in Neural Information Processing Systems, volume 28. Curran Associates, Inc.
- Gabriel Loiseau (3 papers)
- Damien Sileo (27 papers)
- Damien Riquet (3 papers)
- Maxime Meyer (8 papers)
- Marc Tommasi (25 papers)