H. Farid, "Detecting hidden messages using higher-order statistics and support vector machines, " in Proc. of the 5th International Workshop of Information Hiding, (Noordwijkerhout, The Netherlands), Oct. 2002.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....secret message, and even perhaps the steganography key using only the observed stego messages. During this process we exploit spatial diversity and temporal diversity information that will be explained in later sections. Passive steganalysis has been attempted previously by many researchers [7,14,13,11,19,23,21,6,1,27]. The general theme behind these techniques is the exploitation of first and higher order statistics depending on the steganography technique. A priori spatial and frequency domain information about the stego messages are used to arrive at a steganalysis strategy. When such a priori information is ....

S. Lyu and H. Farid. Detecting hidden messages using higher-order statistics and support vector machines. Proc. 5th International Workshop on Information Hiding, 2002.

.... data hiding methods as well as secure steganographic techniques (at least those intended to be secure) With regard to such secure techniques, we begin with the RS steganalysis of Fridrich, Goljan, and Du [9] PoV analysis of Westfeld and Pfitzmann [10] universal blind detection of Farid [11]; and a quality based method of Avciba, Memon, and Sankur [12] RS steganalysis [9] is designed to detect LSB based steganography. The relative smoothness in a particular bitplane is measured over groups of pixels and changes to this smoothness are observed as bits are flipped (negated) The ....

....to be zero as they are to be one, such embedding changes the relative frequencies of the elements of the pairs. The resulting distributions are unlike those of natural imagery. This approach is also applicable to domains other than the pixel domain. Farid s universal blind detection technique [11] is based on high order statistical models. A statistical feature vector is extracted from the image and input to a linear classifier. The classifier must first be trained. This technique has been shown to work well for LSB based steganography and may provide good results for other techniques. ....

H. Farid, "Detecting Hidden Messages Using Higher-Order Statistical Models", IEEE International Conference on Images Processing, 2002.

No context found.

H. Farid, "Detecting hidden messages using higher-order statistical models," in International Conference on Image Processing, (Rochester, New York), 2002.

....vector which is used to discriminate between images that contain hidden messages and those that do not. 3 Classification From the measured statistics of a training set of images with and without hidden messages, the goal is to determine whether a test image contains a message. In earlier work [6], we performed this classification using a Fisher linear discriminant (FLD) analysis [7, 5] Here a more flexible support vector machine (SVM) classifier is employed [24, 25, 3] We briefly describe, in increasing complexity, three classes of SVMs. The first, linear separable case is ....

....results for JPEG, GIF and TIFF format images. Note that the JPEG classifier generalizes to the di#erent embedding programs not previously seen by the classifier. Also shown in this table are results from classification employing a Fisher linear discriminant analysis used in our earlier work [6]. Both the FLD and SVM classifiers employ a linear separating hyperplane for classification so, as expected, performance is similar across these di#erent classifiers. Shown in Table 2 are classification results for a non linear SVM (using a radial basis kernel function) also implemented using ....

[Article contains additional citation context not shown here]

H. Farid. Detecting hidden messages using higher-order statistical models. In International Conference on Image Processing, page (to appear), Rochester, New York, 2002.

....imperceptible to the human eye. In addition to these results, we have shown in previous work that both a Fisher linear discriminant and a support vector machine are effective in discriminating between natural and steg images across a range of different image formats and embedding programs (see [6, 16] for more details) 3.2. Computer Graphics or Photograph In 1996 the Child Pornography Prevention Act was passed which, in part, prohibited any image that appears to be or conveys the impression of someone under 18 engaged in sexually explicit conduct. This law made illegal computer generated ....

S. Lyu and H. Farid. Detecting hidden messages using higher-order statistics and support vector machines. In 5th International Workshop on Information Hiding, Noordwijkerhout, The Netherlands, 2002.

....imperceptible to the human eye. In addition to these results, we have shown in previous work that both a Fisher linear discriminant and a support vector machine are effective in discriminating between natural and steg images across a range of different image formats and embedding programs (see [6, 16] for more details) 3.2. Computer Graphics or Photograph In 1996 the Child Pornography Prevention Act was passed which, in part, prohibited any image that appears to be or conveys the impression of someone under 18 engaged in sexually explicit conduct. This law made illegal computer generated ....

H. Farid. Detecting hidden messages using higher-order statistical models. In International Conference on Image Processing, Rochester, New York, 2002.

No context found.

H. Farid, "Detecting hidden messages using higher-order statistics and support vector machines, " in Proc. of the 5th International Workshop of Information Hiding, (Noordwijkerhout, The Netherlands), Oct. 2002.

No context found.

H. Farid, "Detecting hidden messages using higher-order statistical models," in Proc. ICIP2002.

No context found.

H. Farid, "Detecting Hidden Messages Using HigherOrder Statistical Models," Proc. Int'l Conf. Image Processing, IEEE Press, 2002.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC