Generalised differential privacy for text document processing

Research output: Chapter in Book/Report/Conference proceedingConference proceeding contributionResearchpeer-review

Abstract

We address the problem of how to “obfuscate” texts by removing stylistic clues which can identify authorship, whilst preserving (as much as possible) the content of the text. In this paper we combine ideas from “generalised differential privacy” and machine learning techniques for text processing to model privacy for text documents. We define a privacy mechanism that operates at the level of text documents represented as “bags-of-words”—these representations are typical in machine learning and contain sufficient information to carry out many kinds of classification tasks including topic identification and authorship attribution (of the original documents). We show that our mechanism satisfies privacy with respect to a metric for semantic similarity, thereby providing a balance between utility, defined by the semantic content of texts, with the obfuscation of stylistic clues. We demonstrate our implementation on a “fan fiction” dataset, confirming that it is indeed possible to disguise writing style effectively whilst preserving enough information and variation for accurate content classification tasks. We refer the reader to our complete paper [15] which contains full proofs and further experimentation details.

LanguageEnglish
Title of host publicationPrinciples of Security and Trust
Subtitle of host publication8th International Conference, POST 2019, Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019, Proceedings
EditorsFlemming Nielson, David Sands
Place of PublicationCham
PublisherSpringer-VDI-Verlag GmbH & Co. KG
Pages123-148
Number of pages26
ISBN (Electronic)9783030171384
ISBN (Print)9783030171377
DOIs
Publication statusPublished - 1 Jan 2019
Event8th International Conference on Principles of Security and Trust, POST 2019 Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019 - Prague, Czech Republic
Duration: 6 Apr 201911 Apr 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11426 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference8th International Conference on Principles of Security and Trust, POST 2019 Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019
CountryCzech Republic
CityPrague
Period6/04/1911/04/19

Fingerprint

Privacy
Learning systems
Semantics
Text processing
Processing
Fans
Machine Learning
Text Processing
Obfuscation
Semantic Similarity
Experimentation
Text
Sufficient
Metric
Demonstrate
Model

Bibliographical note

Copyright the Author(s) 2019. Version archived for private and non-commercial use with the permission of the author/s and according to publisher conditions. For further rights please contact the publisher.

Keywords

  • Author obfuscation
  • Earth Mover’s metric
  • Generalised differential privacy
  • Natural language processing

Cite this

Fernandes, N., Dras, M., & McIver, A. (2019). Generalised differential privacy for text document processing. In F. Nielson, & D. Sands (Eds.), Principles of Security and Trust: 8th International Conference, POST 2019, Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019, Proceedings (pp. 123-148). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11426 LNCS). Cham: Springer-VDI-Verlag GmbH & Co. KG. https://doi.org/10.1007/978-3-030-17138-4_6
Fernandes, Natasha ; Dras, Mark ; McIver, Annabelle. / Generalised differential privacy for text document processing. Principles of Security and Trust: 8th International Conference, POST 2019, Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019, Proceedings. editor / Flemming Nielson ; David Sands. Cham : Springer-VDI-Verlag GmbH & Co. KG, 2019. pp. 123-148 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{de45936660fb44f4a9ae286ca45be95c,
title = "Generalised differential privacy for text document processing",
abstract = "We address the problem of how to “obfuscate” texts by removing stylistic clues which can identify authorship, whilst preserving (as much as possible) the content of the text. In this paper we combine ideas from “generalised differential privacy” and machine learning techniques for text processing to model privacy for text documents. We define a privacy mechanism that operates at the level of text documents represented as “bags-of-words”—these representations are typical in machine learning and contain sufficient information to carry out many kinds of classification tasks including topic identification and authorship attribution (of the original documents). We show that our mechanism satisfies privacy with respect to a metric for semantic similarity, thereby providing a balance between utility, defined by the semantic content of texts, with the obfuscation of stylistic clues. We demonstrate our implementation on a “fan fiction” dataset, confirming that it is indeed possible to disguise writing style effectively whilst preserving enough information and variation for accurate content classification tasks. We refer the reader to our complete paper [15] which contains full proofs and further experimentation details.",
keywords = "Author obfuscation, Earth Mover’s metric, Generalised differential privacy, Natural language processing",
author = "Natasha Fernandes and Mark Dras and Annabelle McIver",
note = "Copyright the Author(s) 2019. Version archived for private and non-commercial use with the permission of the author/s and according to publisher conditions. For further rights please contact the publisher.",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-030-17138-4_6",
language = "English",
isbn = "9783030171377",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer-VDI-Verlag GmbH & Co. KG",
pages = "123--148",
editor = "Flemming Nielson and David Sands",
booktitle = "Principles of Security and Trust",
address = "Germany",

}

Fernandes, N, Dras, M & McIver, A 2019, Generalised differential privacy for text document processing. in F Nielson & D Sands (eds), Principles of Security and Trust: 8th International Conference, POST 2019, Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11426 LNCS, Springer-VDI-Verlag GmbH & Co. KG, Cham, pp. 123-148, 8th International Conference on Principles of Security and Trust, POST 2019 Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019, Prague, Czech Republic, 6/04/19. https://doi.org/10.1007/978-3-030-17138-4_6

Generalised differential privacy for text document processing. / Fernandes, Natasha; Dras, Mark; McIver, Annabelle.

Principles of Security and Trust: 8th International Conference, POST 2019, Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019, Proceedings. ed. / Flemming Nielson; David Sands. Cham : Springer-VDI-Verlag GmbH & Co. KG, 2019. p. 123-148 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11426 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference proceeding contributionResearchpeer-review

TY - GEN

T1 - Generalised differential privacy for text document processing

AU - Fernandes, Natasha

AU - Dras, Mark

AU - McIver, Annabelle

N1 - Copyright the Author(s) 2019. Version archived for private and non-commercial use with the permission of the author/s and according to publisher conditions. For further rights please contact the publisher.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - We address the problem of how to “obfuscate” texts by removing stylistic clues which can identify authorship, whilst preserving (as much as possible) the content of the text. In this paper we combine ideas from “generalised differential privacy” and machine learning techniques for text processing to model privacy for text documents. We define a privacy mechanism that operates at the level of text documents represented as “bags-of-words”—these representations are typical in machine learning and contain sufficient information to carry out many kinds of classification tasks including topic identification and authorship attribution (of the original documents). We show that our mechanism satisfies privacy with respect to a metric for semantic similarity, thereby providing a balance between utility, defined by the semantic content of texts, with the obfuscation of stylistic clues. We demonstrate our implementation on a “fan fiction” dataset, confirming that it is indeed possible to disguise writing style effectively whilst preserving enough information and variation for accurate content classification tasks. We refer the reader to our complete paper [15] which contains full proofs and further experimentation details.

AB - We address the problem of how to “obfuscate” texts by removing stylistic clues which can identify authorship, whilst preserving (as much as possible) the content of the text. In this paper we combine ideas from “generalised differential privacy” and machine learning techniques for text processing to model privacy for text documents. We define a privacy mechanism that operates at the level of text documents represented as “bags-of-words”—these representations are typical in machine learning and contain sufficient information to carry out many kinds of classification tasks including topic identification and authorship attribution (of the original documents). We show that our mechanism satisfies privacy with respect to a metric for semantic similarity, thereby providing a balance between utility, defined by the semantic content of texts, with the obfuscation of stylistic clues. We demonstrate our implementation on a “fan fiction” dataset, confirming that it is indeed possible to disguise writing style effectively whilst preserving enough information and variation for accurate content classification tasks. We refer the reader to our complete paper [15] which contains full proofs and further experimentation details.

KW - Author obfuscation

KW - Earth Mover’s metric

KW - Generalised differential privacy

KW - Natural language processing

UR - http://www.scopus.com/inward/record.url?scp=85064891352&partnerID=8YFLogxK

UR - http://purl.org/au-research/grants/arc/DP140101119

U2 - 10.1007/978-3-030-17138-4_6

DO - 10.1007/978-3-030-17138-4_6

M3 - Conference proceeding contribution

SN - 9783030171377

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 123

EP - 148

BT - Principles of Security and Trust

A2 - Nielson, Flemming

A2 - Sands, David

PB - Springer-VDI-Verlag GmbH & Co. KG

CY - Cham

ER -

Fernandes N, Dras M, McIver A. Generalised differential privacy for text document processing. In Nielson F, Sands D, editors, Principles of Security and Trust: 8th International Conference, POST 2019, Held as Part of the European Joint Conferences on Theory and Practice of Software, ETAPS 2019, Proceedings. Cham: Springer-VDI-Verlag GmbH & Co. KG. 2019. p. 123-148. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-17138-4_6