图像隐写的大数据分析方法  

Big data analysis for image steganography

作　　者：田永波 易军凯 TIAN Yongbo;YI Junkai(School of Automation,Beijing Information Science&Technology University,Beijing 100192,China)

机构地区：[1]北京信息科技大学自动化学院,北京100192

出　　处：《北京信息科技大学学报（自然科学版）》2024年第5期81-87,共7页Journal of Beijing Information Science and Technology University（Science and Technology Edition）

基　　金：国家自然科学基金项目(U1636208)。

摘　　要：针对现有图像隐写分析方法存在的通用性较差、检测准确率欠佳等问题,提出了一种图像隐写的大数据分析方法。首先,根据自适应隐写算法的特性,截取待测图像的高复杂度区域作为核心分析图像。其次,利用数据库完成图像匹配,并得到匹配图像与核心分析图像间的隐写差异特征。最后,将特征输入卷积神经网络模型,完成隐写算法的检测与分类。实验结果表明,在图像全部匹配成功的条件下,该方法对6种隐写算法检测的平均准确率达到了93.98%,同时支持空域和频域的图像,具有较强的通用性。To solve the problems of existing image steganography analysis methods,such as poor generalizability and poor detection accuracy,a big data analysis method for image steganography was proposed.Firstly,according to the characteristics of the adaptive steganography algorithms,the high complexity region of the image to be tested was intercepted as the core analysis image.Secondly,the image matching was completed using database to get the steganographic difference features between the matched image and the core analysis image.Finally,the features were input into the convolutional neural network model to complete the detection and classification of the steganography algorithm.The experimental results show that the average accuracy of this method for the detection of six steganography algorithms reaches 93.98%under the condition of successful matching by default.The method supports images in both the spatial and the frequency domain,and possesses strong versatility.

关 键 词：图像隐写分析 图像复杂度 卷积神经网络 隐写术 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]