@inproceedings{4d62cb8d63884380a558496b90a357e0,
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",
address = "Germany",
}