基于特征函数和质量因子的JPEG图像隐写分析

JPEG image steganalysis based on characteristic function and quality factor

提出了一种新的JPEG通用隐写分析方法.通过分析JPEG图像DCT域和小波域的量化噪声模型和隐写模型,发现了量化噪声和嵌入噪声对图像的作用原理,并用直方图特征函数来区分2种噪声对图像的影响.基于JPEG质量因子对图像进行分类,为每个图像分类单独训练支持向量机分类器.实验显示,本文方法相比已有的一些常用JPEG通用隐写分析方法具有更好的检测性能.

In this paper, we proposed a new JPEG universal steganalytic scheme. By analyzing JPEG quantization model and steganographic model in DCT domain and wavelet domain, the study shows difference of quantization noise and embedding noise. Features calculated from the characteristic functions of JPEG individual histograms and wavelet subbands histograms are used to discriminate them. We categorize JPEG images based on quality factor and train different SVM classifier for individual categories. Experiments show that our method has better performance than some exiting JPEG universal steganalytic methods.