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 proceedingConference 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 languageEnglish (US)
    Title of host publicationDigital Forensics and Watermarking - 16th International Workshop, IWDW 2017, Proceedings
    EditorsYun-Qing Shi, Hyoung Joong Kim, Christian Kraetzer, Jana Dittmann
    PublisherSpringer Verlag
    Pages263-272
    Number of pages10
    ISBN (Print)9783319641843
    DOIs
    StatePublished - 2017
    Event16th International Workshop on Digital Forensics and Watermarking, IWDW 2017 - Magdeburg, Germany
    Duration: Aug 23 2017Aug 25 2017

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume10431 LNCS
    ISSN (Print)0302-9743
    ISSN (Electronic)1611-3349

    Other

    Other16th International Workshop on Digital Forensics and Watermarking, IWDW 2017
    CountryGermany
    CityMagdeburg
    Period8/23/178/25/17

    Keywords

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

    ASJC Scopus subject areas

    • Theoretical Computer Science
    • Computer Science(all)

    Fingerprint Dive into the research topics of 'Steganalysis based on awareness of selection-channel and deep learning'. Together they form a unique fingerprint.

  • 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