Steganalysis based on awareness of selection-channel and deep learning

Document Type

Conference Proceeding

Publication Date

1-1-2017

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.

Identifier

85028448682 (Scopus)

ISBN

[9783319641843]

Publication Title

Lecture Notes in Computer Science Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics

External Full Text Location

https://doi.org/10.1007/978-3-319-64185-0_20

e-ISSN

16113349

ISSN

03029743

First Page

263

Last Page

272

Volume

10431 LNCS

Grant

61379155

Fund Ref

National Natural Science Foundation of China

This document is currently not available here.

Share

COinS