视频帧内帧间编码通用隐写分析卷积神经网络  

Video Steganalysis Convolutional Neural Network for Intra and Inter Coding Steganography

作　　者：刘鹏[1] 李松斌[2] LIU Peng;LI Songbin(Haikou Laboratory,Institute of Acoustics,Chinese Academy of Sciences,Haikou,570105,China;National Network New Media Research Center,Institute of Acoustics,Chinese Academy of Sciences,Beijing,100190,China))

机构地区：[1]中国科学院声学研究所南海研究站,海口570105 [2]中国科学院声学研究所国家网络新媒体工程技术研究中心,北京100190

出　　处：《网络新媒体技术》2018年第6期13-16,共4页Network New Media Technology

基　　金：国家自然科学基金(编号:U1636113);海南省重大科技计划项目(编号:ZDKJ201807);海南省自然科学基金(编号:618QN309);中国科学院声学研究所青年英才计划项目(编号:QNYC201829;QNYC201747);中国科学院声学研究所南海研究站资助科研基金项目

摘　　要：基于数据驱动学习模式的深度学习方法已在计算机视觉、语义分析,语音识别以及自然语言处理等众多机器学习相关应用领域取得了成功的应用,并颠覆了这些领域基于"人工特征"的传统范式。受此启发,本文提出了一种基于深度学习的视频帧内帧间编码通用隐写分析方法。由于视频帧内帧间编码信息隐藏本质上都是修改了视频解码帧图像像素值,因此本文从图像域的角度出发,设计了一种视频隐写分析卷积神经网络,将特征提取和分类模块整合到一个可训练的网络模型框架下,以数据驱动的形式自动学习特征并实现分类。实验结果表明,本文方法具有优异的通用隐写检测性能。The data - driven based deep learning method has been successfully applied to many fileds such as computer vision,semantic analysis,speech recognition and natural language processing.It has overturned the traditional paradigm based on "artificial features"in these fields.Inspired by this,we propose a general steganalysis method for both intra and inter coding'steganography based on deep learning.Since the process of intra and inter coding steganography is eventually reflected in the modification of pixel values in decoded frames,we design a video steganalysis network from the perspective of the spatial domain.Feature extraction and classification modules are integrated into a unified and trainable network framework.It automatically learns features and implements classification in a data - driven manner.Experimental results show that the proposed method has good performance for the detection of intra and inter coding steganography.

关 键 词：卷积神经网络 信息隐藏 隐写分析 视频 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]