基于拉普拉斯模型的序列隐写密钥估计

图像序列隐写是指利用载体图像特征数据（包括频域数据、空域数据）连续嵌入信息的隐藏方法。本文提出了一种针对图像扩展频谱序列隐写的密钥估计算法。该方法基于序列分析与突变检测的理论，利用序列概率比累积和检验方法（CUSUM-SPRT）对变化进行检测。考虑图像DCT系数满足拉普拉斯分布，给出了理想平稳拉普拉斯分布信号扩展频谱隐藏密钥估计的模型。采用随机微分方程法（SDE）生成拉氏分布的随机序列进行实验。对于非平稳信号的图像数据，在低信噪比（SNR）下，利用当地最有效序列检测法，给出了拉普拉斯分布的密钥估计模型。实验显示，该方法不但能检测出扩展频谱隐写，估计嵌入密钥，而且比Trivedi的方法更有效。

Secret Key Estimation in Sequential Steganography Based on the Laplacian Model

Image sequential steganography is defined as a class of embedding algorithms that hide messages in the consecutive features(time,spatial or frequency domain)of the host signal.This paper presents a steganalysis method that estimates the secret key used in the sequential steganography of spread spectrum embedding.We use the method of cumulative sum based on the sequential probability test(CUMUM-SPRT)to detect the changes,which is used in sequential analysis and abrupt changes detection.Considering the Laplacian distribution of image DCT coefficients,we give a secret key estimation model of stationary Laplacian host signals embedded by the spread spectrum steganography.Experiments are done using the stationary Laplacian random sequence generated by a stochastic differential equation(SDE)method.For non-statioanry digital images with a low signal noise ratio(SNR),a locally most powerful steganaysis detector is derived based on the Laplacian distribution.The results of experiments show that our method is more efficient than Trivedi's method.