一种保持直方图特性的JPEG图像掩密算法

A Steganography Preserving the Property of the Histogram for JPEG Images

传统的JPEG图像掩密算法往往因为改变了载体图像的DCT系数直方图而不能抵抗各种基于直方图的攻击,或者为了矫正直方图而降低了隐藏容量,该文提出一种新的保持直方图特性的算法。该算法在嵌入秘密消息时,动态建立Adjust_info表记录系数直方图被改变的情况,然后根据该表调整系数的变化方向,这样不但可以补偿直方图的改变,而且并不降低隐藏容量。对多幅图像的实验表明,该算法不但能够达到13％的高隐藏容量(压缩因子为75),而且掩密后图像总体DCT系数直方图的平均失真率为0．81％。由改进算法掩密后图像低频DCT系数直方图的最大失真率为3．83％,平均失真率只有1．98％,仅对直方图做了微小的改变,因此可以有效抵御各种直方图攻击,从而为安全掩密提供了可靠保证。

Traditional steganographic techniques for JPEG images always suffer from histogram-based attacks because they change the DCT coefficient histogram of the cover image, or from the reduced hiding capacity due to the correction of the change of the histogram. A new steganography preserving the property of the histogram is proposed in this paper. When the secret message is embedded, an adjust_info table is built to memorize the alteration of the histogram. Then the modification direction of the coefficient is adjusted according to the table. In this way, the change of the histogram is compensated and the hiding capacity is not reduced yet. Several experimental results demonstrate that the algorithm can attain a high hiding capacity of 13% （the quality factor is 75）. Furthermore the average distortion rate of the general DCT coefficient histograms on stego＋images is 0.81%, the maximal distortion rate of the histograms on low frequency coefficients is 3.83% obtained from the stego-images generated by the improved algorithm and the average distortion rate of them is only 1.98%. The property of the histogram is fine changed; therefore, various histogram-based attacks can be effectively resisted and secure steganorgraphy can be insured.