Evaluation of deep neural network architectures for authorship obfuscation of Portuguese texts

Preserving authorship anonymity is paramount to protect activists, freedom of expression, and critical journalism. Although there are several mechanisms to provide anonymity on the Internet, one can still identify anonymous authors through their writing style. With the advances in neural network and...

Full description

Saved in:
Bibliographic Details
Main Authors: Antônio Marcos Rodrigues Franco, Ítalo Cunha, Leonardo B. Oliveira
Format: Article
Language:English
Published: Elsevier 2024-12-01
Series:Natural Language Processing Journal
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2949719124000554
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846123181315719168
author Antônio Marcos Rodrigues Franco
Ítalo Cunha
Leonardo B. Oliveira
author_facet Antônio Marcos Rodrigues Franco
Ítalo Cunha
Leonardo B. Oliveira
author_sort Antônio Marcos Rodrigues Franco
collection DOAJ
description Preserving authorship anonymity is paramount to protect activists, freedom of expression, and critical journalism. Although there are several mechanisms to provide anonymity on the Internet, one can still identify anonymous authors through their writing style. With the advances in neural network and natural language processing research, the success of a classifier when identifying the author of a text is growing. On the other hand, new approaches that use recurrent neural networks for automatic generation of obfuscated texts have also arisen to fight anonymity adversaries. In this work, we evaluate two approaches that use neural networks to generate obfuscated texts. The first approach uses Generative Adversarial Networks to train an encoder–decoder to transform sentences from an input style into a target style. The second one trains an auto encoder with Gradient Reversal Layer to learn invariant representations. In our experiments, we compared the efficiency of both techniques when removing the stylistic attributes of a text and preserving its original semantics. Our evaluation on real texts clarifies each technique’s trade-offs for Portuguese texts and provides guidance on practical deployment.
format Article
id doaj-art-3b563a19b0154d19821ce267ccf7a865
institution Kabale University
issn 2949-7191
language English
publishDate 2024-12-01
publisher Elsevier
record_format Article
series Natural Language Processing Journal
spelling doaj-art-3b563a19b0154d19821ce267ccf7a8652024-12-14T06:34:32ZengElsevierNatural Language Processing Journal2949-71912024-12-019100107Evaluation of deep neural network architectures for authorship obfuscation of Portuguese textsAntônio Marcos Rodrigues Franco0Ítalo Cunha1Leonardo B. Oliveira2Universidade Federal de Minas Gerais, Av. Presidente Antonio Carlos, 6627, Belo Horizonte, 31270010, Minas Gerais, BrazilCorresponding author.; Universidade Federal de Minas Gerais, Av. Presidente Antonio Carlos, 6627, Belo Horizonte, 31270010, Minas Gerais, BrazilUniversidade Federal de Minas Gerais, Av. Presidente Antonio Carlos, 6627, Belo Horizonte, 31270010, Minas Gerais, BrazilPreserving authorship anonymity is paramount to protect activists, freedom of expression, and critical journalism. Although there are several mechanisms to provide anonymity on the Internet, one can still identify anonymous authors through their writing style. With the advances in neural network and natural language processing research, the success of a classifier when identifying the author of a text is growing. On the other hand, new approaches that use recurrent neural networks for automatic generation of obfuscated texts have also arisen to fight anonymity adversaries. In this work, we evaluate two approaches that use neural networks to generate obfuscated texts. The first approach uses Generative Adversarial Networks to train an encoder–decoder to transform sentences from an input style into a target style. The second one trains an auto encoder with Gradient Reversal Layer to learn invariant representations. In our experiments, we compared the efficiency of both techniques when removing the stylistic attributes of a text and preserving its original semantics. Our evaluation on real texts clarifies each technique’s trade-offs for Portuguese texts and provides guidance on practical deployment.http://www.sciencedirect.com/science/article/pii/S2949719124000554Authorship obfuscationPrivacyNatural language processingArtificial intelligence
spellingShingle Antônio Marcos Rodrigues Franco
Ítalo Cunha
Leonardo B. Oliveira
Evaluation of deep neural network architectures for authorship obfuscation of Portuguese texts
Natural Language Processing Journal
Authorship obfuscation
Privacy
Natural language processing
Artificial intelligence
title Evaluation of deep neural network architectures for authorship obfuscation of Portuguese texts
title_full Evaluation of deep neural network architectures for authorship obfuscation of Portuguese texts
title_fullStr Evaluation of deep neural network architectures for authorship obfuscation of Portuguese texts
title_full_unstemmed Evaluation of deep neural network architectures for authorship obfuscation of Portuguese texts
title_short Evaluation of deep neural network architectures for authorship obfuscation of Portuguese texts
title_sort evaluation of deep neural network architectures for authorship obfuscation of portuguese texts
topic Authorship obfuscation
Privacy
Natural language processing
Artificial intelligence
url http://www.sciencedirect.com/science/article/pii/S2949719124000554
work_keys_str_mv AT antoniomarcosrodriguesfranco evaluationofdeepneuralnetworkarchitecturesforauthorshipobfuscationofportuguesetexts
AT italocunha evaluationofdeepneuralnetworkarchitecturesforauthorshipobfuscationofportuguesetexts
AT leonardoboliveira evaluationofdeepneuralnetworkarchitecturesforauthorshipobfuscationofportuguesetexts