基于ICA模型的主动隐写分析

Active Steganalysis Based on ICA Model

提出一种基于独立成分分析(ICA)模型的主动隐写分析方案。该方案假设秘密信息是独立同分布序列且统计独立于载体图像,将隐写分析过程视为ICA模型的求解问题。借助于最大后验概率估计器,该方案仅使用一幅隐写图像即能提取出秘密信息,克服了Chandramouli所提方案的局限性。仿真实验结果表明,该方案能提取大约80%的秘密信息,且性能随嵌入长度的增加而提高。

A new active steganalysis scheme which only uses one copy of stego image is presented. This paper views active steganalysis as an Independent Component Analysis（ICA） problem under the assumption that embedded secret message is an independent, identically distributed （i.i.d） random sequence and independent to cover image. With only one copy of stego image, it adoptes Maximum Posteriori（MAP） estimator to obtain an estimation of cover image and generates another two signals with the estimated version. All the three signals are as input signals of ICA algorithm. The proposed scheme overcomes the constraint of Chandramouli＇s method which needs two copies of stego image. Experimental results show that the proposed method achieves acceptable performance and improves its performance with larger message length.