摘要
针对现有网络隐写分析算法特征提取难度大、算法适用范围单一的问题,文章提出了一种基于卷积神经网络的网络隐写分析方法。对网络数据流进行预处理,将所有数据包处理成大小相同的矩阵,最大限度地保留数据特征完整性;使用异构卷积进行特征提取,减少模型计算量及参数数量,加快模型收敛速度;取消池化层,提高模型训练效率。与传统网络隐写分析方法相比,模型能够自动提取数据特征,识别多种网络隐写算法。
In terms of the difficulties in feature extraction and single application scope of existing network steganalysis algorithms,a network steganalysis method based on convolutional neural network is proposed in this paper.The model preprocessed the network data stream and makes all packets into a matrix of the same size to preserve the integrity of data feature to the maximum extent.The model uses heterogeneous convolution for feature extraction,reducing the amount of model calculation and parameters,and speeding up the model convergence speed;pooling layer is canceled to improve the model training efficiency.Compared with the traditional network steganalysis method,the model can automatically extract data features and identify multiple network steganography algorithms.
作者
赵丹阳
Zhao Danyang(College of Computer Science&Technology,Qingdao University,Qingdao 266000,China)
出处
《无线互联科技》
2020年第6期36-37,44,共3页
Wireless Internet Technology
关键词
网络隐写分析
卷积神经网络
特征提取
network steganalysis
convolutional neural network
feature extraction
