Steganalysis based on awareness of selection-channel and deep learning

Jianhua Yang; Kai Liu; Xiangui Kang; Edward Wong; Yunqing Shi

doi:10.1007/978-3-319-64185-0_20

Steganalysis based on awareness of selection-channel and deep learning

Jianhua Yang, Kai Liu, Xiangui Kang, Edward Wong, Yunqing Shi

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Recently, deep learning has been used in steganalysis based on convolutional neural networks (CNN). In this work, we propose a CNN architecture (the so-called maxCNN) to use the selection channel. It is the first time that the knowledge of the selection channel has been incorporated into CNN for steganalysis. The proposed method assigns large weights to features learned from complex texture regions while assigns small weights to features learned from smooth regions. Experimental results on the well-known dataset BOSS-base have demonstrated that the proposed scheme is able to improve detection performance, especially for low embedding payloads. The results have shown that with the ensemble of maxCNN and maxSRMd2+EC, the proposed method can obtain better performance compared with the reported state-of-the-art on detecting WOW embedding algorithm.

Original language	English (US)
Title of host publication	Digital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings
Editors	Yun-Qing Shi, Hyoung Joong Kim, Christian Kraetzer, Jana Dittmann
Publisher	Springer Verlag
Pages	263-272
Number of pages	10
ISBN (Print)	9783319641843
DOIs	https://doi.org/10.1007/978-3-319-64185-0_20
State	Published - 2017
Event	16th International Workshop on Digital Forensics and Watermarking, IWDW 2017 - Magdeburg, Germany Duration: Aug 23 2017 → Aug 25 2017

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10431 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	16th International Workshop on Digital Forensics and Watermarking, IWDW 2017
Country/Territory	Germany
City	Magdeburg
Period	8/23/17 → 8/25/17

Keywords

Adaptive steganography
Convolutional neural networks (CNN)
Selection-channel
Steganalysis

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-319-64185-0_20

Cite this

Yang, J., Liu, K., Kang, X., Wong, E., & Shi, Y. (2017). Steganalysis based on awareness of selection-channel and deep learning. In Y.-Q. Shi, H. J. Kim, C. Kraetzer, & J. Dittmann (Eds.), Digital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings (pp. 263-272). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10431 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-64185-0_20

Steganalysis based on awareness of selection-channel and deep learning. / Yang, Jianhua; Liu, Kai; Kang, Xiangui et al.
Digital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings. ed. / Yun-Qing Shi; Hyoung Joong Kim; Christian Kraetzer; Jana Dittmann. Springer Verlag, 2017. p. 263-272 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10431 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Yang, J, Liu, K, Kang, X, Wong, E & Shi, Y 2017, Steganalysis based on awareness of selection-channel and deep learning. in Y-Q Shi, HJ Kim, C Kraetzer & J Dittmann (eds), Digital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10431 LNCS, Springer Verlag, pp. 263-272, 16th International Workshop on Digital Forensics and Watermarking, IWDW 2017, Magdeburg, Germany, 8/23/17. https://doi.org/10.1007/978-3-319-64185-0_20

Yang J, Liu K, Kang X, Wong E, Shi Y. Steganalysis based on awareness of selection-channel and deep learning. In Shi YQ, Kim HJ, Kraetzer C, Dittmann J, editors, Digital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings. Springer Verlag. 2017. p. 263-272. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-64185-0_20

Yang, Jianhua ; Liu, Kai ; Kang, Xiangui et al. / Steganalysis based on awareness of selection-channel and deep learning. Digital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings. editor / Yun-Qing Shi ; Hyoung Joong Kim ; Christian Kraetzer ; Jana Dittmann. Springer Verlag, 2017. pp. 263-272 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{52c8b6bdd3dc4ad48780e8796b865d8e,

title = "Steganalysis based on awareness of selection-channel and deep learning",

abstract = "Recently, deep learning has been used in steganalysis based on convolutional neural networks (CNN). In this work, we propose a CNN architecture (the so-called maxCNN) to use the selection channel. It is the first time that the knowledge of the selection channel has been incorporated into CNN for steganalysis. The proposed method assigns large weights to features learned from complex texture regions while assigns small weights to features learned from smooth regions. Experimental results on the well-known dataset BOSS-base have demonstrated that the proposed scheme is able to improve detection performance, especially for low embedding payloads. The results have shown that with the ensemble of maxCNN and maxSRMd2+EC, the proposed method can obtain better performance compared with the reported state-of-the-art on detecting WOW embedding algorithm.",

keywords = "Adaptive steganography, Convolutional neural networks (CNN), Selection-channel, Steganalysis",

author = "Jianhua Yang and Kai Liu and Xiangui Kang and Edward Wong and Yunqing Shi",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing AG 2017.; 16th International Workshop on Digital Forensics and Watermarking, IWDW 2017 ; Conference date: 23-08-2017 Through 25-08-2017",

year = "2017",

doi = "10.1007/978-3-319-64185-0_20",

language = "English (US)",

isbn = "9783319641843",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "263--272",

editor = "Yun-Qing Shi and Kim, {Hyoung Joong} and Christian Kraetzer and Jana Dittmann",

booktitle = "Digital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings",

}

TY - GEN

T1 - Steganalysis based on awareness of selection-channel and deep learning

AU - Yang, Jianhua

AU - Liu, Kai

AU - Kang, Xiangui

AU - Wong, Edward

AU - Shi, Yunqing

N1 - Publisher Copyright: © Springer International Publishing AG 2017.

PY - 2017

Y1 - 2017

N2 - Recently, deep learning has been used in steganalysis based on convolutional neural networks (CNN). In this work, we propose a CNN architecture (the so-called maxCNN) to use the selection channel. It is the first time that the knowledge of the selection channel has been incorporated into CNN for steganalysis. The proposed method assigns large weights to features learned from complex texture regions while assigns small weights to features learned from smooth regions. Experimental results on the well-known dataset BOSS-base have demonstrated that the proposed scheme is able to improve detection performance, especially for low embedding payloads. The results have shown that with the ensemble of maxCNN and maxSRMd2+EC, the proposed method can obtain better performance compared with the reported state-of-the-art on detecting WOW embedding algorithm.

AB - Recently, deep learning has been used in steganalysis based on convolutional neural networks (CNN). In this work, we propose a CNN architecture (the so-called maxCNN) to use the selection channel. It is the first time that the knowledge of the selection channel has been incorporated into CNN for steganalysis. The proposed method assigns large weights to features learned from complex texture regions while assigns small weights to features learned from smooth regions. Experimental results on the well-known dataset BOSS-base have demonstrated that the proposed scheme is able to improve detection performance, especially for low embedding payloads. The results have shown that with the ensemble of maxCNN and maxSRMd2+EC, the proposed method can obtain better performance compared with the reported state-of-the-art on detecting WOW embedding algorithm.

KW - Adaptive steganography

KW - Convolutional neural networks (CNN)

KW - Selection-channel

KW - Steganalysis

UR - http://www.scopus.com/inward/record.url?scp=85028448682&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85028448682&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-64185-0_20

DO - 10.1007/978-3-319-64185-0_20

M3 - Conference contribution

AN - SCOPUS:85028448682

SN - 9783319641843

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

SP - 263

EP - 272

BT - Digital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings

A2 - Shi, Yun-Qing

A2 - Kim, Hyoung Joong

A2 - Kraetzer, Christian

A2 - Dittmann, Jana

PB - Springer Verlag

T2 - 16th International Workshop on Digital Forensics and Watermarking, IWDW 2017

Y2 - 23 August 2017 through 25 August 2017

ER -

Steganalysis based on awareness of selection-channel and deep learning

Abstract

Publication series

Other

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this