LinL:Lost in n-best list

  • Peng Meng*
  • , Yun Qing Shi
  • , Liusheng Huang
  • , Zhili Chen
  • , Wei Yang
  • , Abdelrahman Desoky
  • *Corresponding author for this work

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

10 Scopus citations

Abstract

Translation-based steganography (TBS) is a new kind of text steganographic scheme. However, contemporary TBS methods are vulnerable to statistical attacks. Differently, this paper presents a novel TBS, namely Lost in n-best List, abbreviated as LinL, that is resilient against the current statistical attacks. LinL employs only one Statistical Machine Translator (SMT) in the encoding process which selects one of the n-best list of each cover text sentence in order to camouflage messages in stegotext. The presented theoretical analysis demonstrates that there is a classification accuracy upper bound between normal translated text and the stegotext. When the text size is 1000 sentences, the theoretical maximum classification accuracy is about 60%. The experiment results also show current steganalysis methods cannot detect LinL.

Original languageEnglish
Title of host publicationInformation Hiding - 13th International Conference, IH 2011, Revised Selected Papers
Pages329-341
Number of pages13
DOIs
StatePublished - 2011
Externally publishedYes
Event13th International Conference on Information Hiding, IH 2011 - Prague, Czech Republic
Duration: 18 May 201120 May 2011

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume6958 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference13th International Conference on Information Hiding, IH 2011
Country/TerritoryCzech Republic
CityPrague
Period18/05/1120/05/11

Keywords

  • LinL
  • linguistic steganography
  • natural language steganography
  • text steganography
  • translation-based steganography (TBS)

Fingerprint

Dive into the research topics of 'LinL:Lost in n-best list'. Together they form a unique fingerprint.

Cite this