基于模糊积分多分类器融合的JPEG 图像隐写算法识别

提取和恢复嵌入信息是隐写分析的最终目的,嵌入算法识别是秘密信息提取的前提。目前研究者针对隐写算法识别提出了多种隐写分析系统,但这些系统各有优缺点,因此实际应用中难以取舍。文章提出基于模糊积分的多分类器融合方法来识别JPEG图像隐写算法。实验结果表明,相对于单个隐写分析系统,通过分类器融合后识别精度提高约5%。

Multi-Class Classification Fusion Using Fuzzy Integral for Identifying JPEG Steganography Methods

The extraction and recovery of hidden information is the aim of steganalysis,and the premise step in extraction is to identify the embedding method.Several steganalysis systems have been developed for this purpose,but each system has its advantages and defects so it＇s difficult to determine the detection system for identifying the embedding method correctly.This paper applies fuzzy integral to fuse multiple steganalysis systems and identifies the embedding method used to create a stego JPEG image.Experimental results indicate that through the novel addition of the classifier fusion step to the multi-class steganalysis system,the classification accuracy is improved by about 5% compared with the individual steganalysis system.