基于EM算法的数字取证中图像重采样检测.pdf

基于 EM 算法的数字取证中图像重采样检测 朱秀明 1 姚秋明 1 宣国荣 1 童学锋 1 施云庆 2 1、同济大学 计算机系 上海 2 、新泽西理工学院 计算机与电气工程系 美国 (xiumingzhu@) 摘要：指出了数字取证中检测图象重采样痕迹，实质上是发现图象中的周期信息。本文对已 有的模型进行了改进，将原始图像分布代替均匀分布并增加了先验概率的迭代，使得分布情 况更符合实际。利用编程技巧解决了在使用 EM 算法迭代收敛的过程中容易奇异的问题，保 证算法成功收敛。实验证实了此算法的有效性：对于未压缩图像的缩小、放大和旋转，此算 法可以检测出来。同时，也可以检测精确的检测出图像是否 JPEG 压缩过。 关键字：EM 算法；数字取证；重采样；参数估计 Re-Sampling Detection in Digital Forensics based on EM Algorithm 1 1 1 1 1 Xiuming Zhu Qiuming Yao Guorong Xuan Xuefeng Tong Yun Q. Shi 1. Dept. of Computer Science, Tongji University, China 2. Dept. of Electrical and Computer Engineering ,New Jersey Institute of Technology, USA (xiumingzhu@) Abstract: The re-sampling is often used in image tampering, e.g., in scaling or rotating. Hence the re-sampling detection plays an important role in digital forensics. The EM algorithm is recently used in the re-sampling detection. In this study, the used EM algorithm has been modified and improved. The possible singular points during the EM iterations are avoided by special procedures in software implementation. A priori probability updating is added in order to approach the realistic distribution. Experimental investigations on re-sapling detection have demonstrated the effectiveness of the modified EM in detecting image re-sampling. For one thing, it can detect re-sampling traces on uncompressed bmp or tiff images. For another, it even can detect whether an image is JPEG compressed or not and re-sampling traces on JPEG images under certain situation. keywords: EM algorithm;Forensics