基于支持向量数据描述和不确定性推理的单类隐写分析算法

One-class Steganalysis Method Based on Support Vector Data Description and Uncertainty Reasoning

单类隐写分析较传统的二类隐写分析有更好的"盲"检测特性和更强的通用性。针对单类隐写分析可靠性较低的问题,将不确定性推理理论应用于隐写分析中降低不确定性因素对单类隐写分析可靠性的影响,提出了一种基于支持数据描述和不确定性推理的单类通用隐写分析算法。实验表明,算法具有较好的可靠性、鲁棒性、通用性。

The one-class universal steganaylsis algorithm has better performance in terms of blind detection and universality than two-class algorithm.This method has poor accuracy of detection,however,it is necessary to improve the accuracy of one-class steganaylsis algorithm from feature extracting,classifying and detecting.Therefore,the algorithm based on support vector data description and uncertainty reasoning is put forward.In this algorithm,due to support vector description has better generalization capability,it is used as classifier and the probability expression is designed.Moreover,the uncertainty reasoning theory and method are applied to reduce the negative effect of some uncertainty factors in the algorithm.The experimental results show that the algorithm has good reliability,robustness and universality.