توضیحات
ABSTRACT
In this paper, we present a scheme based on feature mining and neuro-fuzzy inference system for detecting LSB matching steganography in grayscale images, which is a very challenging problem in steganalysis. Four types of features are proposed, and a Dynamic Evolving Neural Fuzzy Inference System (DENFIS) based feature selection is proposed, as well as the use of Support Vector Machine Recursive Feature Elimination (SVM-RFE) to obtain better detection accuracy. In comparison with other well-known features, overall, our features perform the best. DENFIS outperforms some traditional learning classifiers. SVM-RFE and DENFIS based feature selection outperform statistical significance based feature selection such as t-test. Experimental results also indicate that it remains very challenging to steganalyze LSB matching steganography in grayscale images with high complexity.
INTRODUCTION
Steganalysis is the science and art of detecting the presence of hidden data in digital images, audios, videos and other media. In steganography or the hiding of secret data in digital media, the most common cover is digital images. To this date, many steganographical or embedding methods, such as LSB embedding, spread spectrum steganography, F5 algorithm and some other JPEG steganography, have been very successfully steganalyzed [Fridrich et al., 2003; Ker, 2005a; Fridrich et al. 2002; Harmsen and Pearlman 2003; Choubassi and Moulin 2005; Liu et al., 2006a]. Several other embedding paradigms, include stochastic modulation [Fridrich and Goljan, 2003; Moulin and Briassouli, 2004] and LSB matching [Sharp 2001], however, are much more difficult to detect. The literature does provide a few detectors for LSB matching steganography. One of the first papers on detection of embedding by noise adding is the paper by Harmsen and Pearlman [Harmsen and Pearlman, 2003], wherein the measure of histogram characteristic function center of mass (HCFCOM), is extracted and a Bayesian multivariate classifier is applied. Based on the contribution of Harmsen and Pearlman [2003], Ker [2005b] proposes two novel ways of applying the HCF: calibrating the output using a down-sampled image and computing the adjacency histogram instead of the usual histogram. The best discriminators are Adjacency HCFCOM (A.HCFCOM) and Calibrated Adjacency HCFCOM (C.A.HCFCOM) to improve the probability of detection for LSB matching in grayscale images. Farid and Lyu describe an approach to detecting hidden messages in images by using a wavelet-like decomposition to build high-order statistical models of natural images [Lyu and Farid, 2004 and 2005]. Fridrich et al. [2005] propose a Maximum Likelihood (ML) estimator for estimating the number of embedding changes for non-adaptive ±K embedding in images. Based on the stego-signal estimation, Holotyak et al. [2005] present a blind steganalysis classifying on high order statistics of the estimation signal.
Year: 2007
Publisher : ICASA,IASP
By : Qingzhong Liu , Andrew H. Sung
File Information: English Language/6 Page / size:123 KB
Download: click
سال : 2007
ناشر : ICASA,IASP
کاری از : Qingzhong Liu , Andrew H. Sung
اطلاعات فایل : زبان انگلیسی / 6 صفحه / حجم : KB 123
لینک دانلود : روی همین لینک کلیک کنید
نقد و بررسیها
هنوز بررسیای ثبت نشده است.