Detecting Hidden Messages Using Higher-Order Statistics and Support Vector Machines Siwei Lyu and Hany Farid Department of Computer Science, Dartmouth.

Slides:



Advertisements
Similar presentations
Chapter 6 Matrix Algebra.
Advertisements

Statistical Machine Learning- The Basic Approach and Current Research Challenges Shai Ben-David CS497 February, 2007.
Support Vector Machine
ETHEM ALPAYDIN © The MIT Press, Lecture Slides for.
11/11/02 IDR Workshop Dealing With Location Uncertainty in Images Hasan F. Ates Princeton University 11/11/02.
University of Ioannina - Department of Computer Science Wavelets and Multiresolution Processing (Background) Christophoros Nikou Digital.
Chapter 3 Image Enhancement in the Spatial Domain.
Chapter 10 Curve Fitting and Regression Analysis
IMAN SAUDY UMUT OGUR NORBERT KISS GEORGE TEPES-NICA BARLEY SEEDS CLASSIFICATION.
The Disputed Federalist Papers : SVM Feature Selection via Concave Minimization Glenn Fung and Olvi L. Mangasarian CSNA 2002 June 13-16, 2002 Madison,
1 A Markov Process Based Approach to Effective Attacking JPEG Steganography By Y. Q. Shi, Chunhua Chen, Wen Chen NJIT Presented by Hanlin Hu and Xiao Zhang.
Feature Vector Selection and Use With Hidden Markov Models to Identify Frequency-Modulated Bioacoustic Signals Amidst Noise T. Scott Brandes IEEE Transactions.
Coefficient Path Algorithms Karl Sjöstrand Informatics and Mathematical Modelling, DTU.
Communication & Multimedia C. -H. Hong 2015/6/12 Contourlet Student: Chao-Hsiung Hong Advisor: Prof. Hsueh-Ming Hang.
Robust Real-time Object Detection by Paul Viola and Michael Jones ICCV 2001 Workshop on Statistical and Computation Theories of Vision Presentation by.
Support Vector Machines Formulation  Solve the quadratic program for some : min s. t.,, denotes where or membership.  Different error functions and measures.
Real-time Combined 2D+3D Active Appearance Models Jing Xiao, Simon Baker,Iain Matthew, and Takeo Kanade CVPR 2004 Presented by Pat Chan 23/11/2004.
Wavelet Transform A very brief look.
Support Vector Machines Kernel Machines
CS443: Digital Imaging and Multimedia Filters Spring 2008 Ahmed Elgammal Dept. of Computer Science Rutgers University Spring 2008 Ahmed Elgammal Dept.
Feature Screening Concept: A greedy feature selection method. Rank features and discard those whose ranking criterions are below the threshold. Problem:
Elastic Registration in the Presence of Intensity Variations Source: IEEE Transactions on Medical Imaging, Vol. 22, No. 7, July 2003 Authors: Senthil Periaswamy,
1 How Realistic is Photorealistic?. 2 Yaniv Lefel Hagay Pollak Based on the work of - Siwei Lyu and Hany Farid.
Engineering Computation Curve Fitting 1 Curve Fitting By Least-Squares Regression and Spline Interpolation Part 7.
The Pinhole Camera Model
Optimization Theory Primal Optimization Problem subject to: Primal Optimal Value:
Normalised Least Mean-Square Adaptive Filtering
Steganography By: Joe Jupin Supervised by: Dr. Longin Jan Latecki.
ENG4BF3 Medical Image Processing
Least-Squares Regression
5.1 Basic Estimation Techniques  The relationships we theoretically develop in the text can be estimated statistically using regression analysis,  Regression.
The Wavelet Tutorial: Part3 The Discrete Wavelet Transform
1 Graph Embedding (GE) & Marginal Fisher Analysis (MFA) 吳沛勳 劉冠成 韓仁智
3/2003 Rev 1 I – slide 1 of 33 Session I Part I Review of Fundamentals Module 2Basic Physics and Mathematics Used in Radiation Protection.
Texture. Texture is an innate property of all surfaces (clouds, trees, bricks, hair etc…). It refers to visual patterns of homogeneity and does not result.
Discriminant Functions
IEEE TRANSSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: Deterministic vs. Random Maximum A Posteriori Maximum Likelihood Minimum.
Machine Learning Using Support Vector Machines (Paper Review) Presented to: Prof. Dr. Mohamed Batouche Prepared By: Asma B. Al-Saleh Amani A. Al-Ajlan.
Steganalysis of audio: attacking the Steghide
Part I: Image Transforms DIGITAL IMAGE PROCESSING.
Feature Vector Selection and Use With Hidden Markov Models to Identify Frequency-Modulated Bioacoustic Signals Amidst Noise T. Scott Brandes IEEE Transactions.
Wavelets and Multiresolution Processing (Wavelet Transforms)
Digital Image Processing Lecture 16: Segmentation: Detection of Discontinuities Prof. Charlene Tsai.
Support Vector Machines Project מגישים : גיל טל ואורן אגם מנחה : מיקי אלעד נובמבר 1999 הטכניון מכון טכנולוגי לישראל הפקולטה להנדסת חשמל המעבדה לעיבוד וניתוח.
Categorization by Learning and Combing Object Parts B. Heisele, T. Serre, M. Pontil, T. Vetter, T. Poggio. Presented by Manish Jethwa.
CSE4334/5334 DATA MINING CSE4334/5334 Data Mining, Fall 2014 Department of Computer Science and Engineering, University of Texas at Arlington Chengkai.
Support Vector Machines Tao Department of computer science University of Illinois.
Subband Coding Jennie Abraham 07/23/2009. Overview Previously, different compression schemes were looked into – (i)Vector Quantization Scheme (ii)Differential.
Digital Image Forensics CS 365 By:- - Abhijit Sarang - Pankaj Jindal.
Sub-Band Coding Multimedia Systems and Standards S2 IF Telkom University.
Digital Image Processing Lecture 16: Segmentation: Detection of Discontinuities May 2, 2005 Prof. Charlene Tsai.
1 Detecting Hidden Messages using higher-order stats and SVMs Siwei Lyu and Hany Farid.
A Parallel Mixture of SVMs for Very Large Scale Problems Ronan Collobert Samy Bengio Yoshua Bengio Prepared : S.Y.C. Neural Information Processing Systems,
1 A Markov Process Based Approach to Effective Attacking JPEG Steganography By Y. Q. Shi, Chunhua Chen, Wen Chen NJIT Presented by Ashish Ratnakar and.
Estimation Econometría. ADE.. Estimation We assume we have a sample of size T of: – The dependent variable (y) – The explanatory variables (x 1,x 2, x.
Spectral Methods for Dimensionality
Image Transformation Spatial domain (SD) and Frequency domain (FD)
Digital Image Processing Lecture 16: Segmentation: Detection of Discontinuities Prof. Charlene Tsai.
Image Pre-Processing in the Spatial and Frequent Domain
Learning with information of features
Exposing Digital Forgeries by Detecting Traces of Resampling Alin C
Josh H. McDermott, Eero P. Simoncelli  Neuron 
Generally Discriminant Analysis
Quadrature-Mirror Filter Bank
Marian Stewart Bartlett, Gwen C. Littlewort, Mark G. Frank, Kang Lee 
Detecting Hidden Message Using Higher Order Statistical Models Hany Farid By Jingyu Ye Yiqi Hu.
Linear Discrimination
SVMs for Document Ranking
A Second Order Statistical Analysis of the 2D Discrete Wavelet Transform Corina Nafornita1, Ioana Firoiu1,2, Dorina Isar1, Jean-Marc Boucher2, Alexandru.
Presentation transcript:

Detecting Hidden Messages Using Higher-Order Statistics and Support Vector Machines Siwei Lyu and Hany Farid Department of Computer Science, Dartmouth College, Hanover, NH 03755, USA IEEE International Conference on Image Processing,2002 1

Outline Introduction Image Statistics Classification Results Conclusion References 2

Introduction Although the presence of embedded messages is often imperceptible to the human eye, it may nevertheless disturb the statistics of an image. The embedding of a message significantly alters these statistics and thus becomes detectable. Support vector machines (linear and non-linear) are employed to detect these statistical deviations. 3

Image Statistics The decomposition employed here is based on separable quadrature mirror filters (QMFs) 4

Image Statistics This is accomplished by applying separable lowpass and highpass filters along the image axes generating a vertical, horizontal, diagonal and lowpass subband. 5

Image Statistics The vertical, horizontal, and diagonal sub bands at scale i = 1,..., n are denoted as V i (x, y), H i (x,y), and D i (x,y), respectively 6

Image Statistics The second set of statistics is based on the errors in an optimal linear predictor of coefficient magnitude 7

Image Statistics Where denotes scalar weighting values. This linear relationship is expressed more compactly in matrix form as The coefficients are determined by minimizing the quadratic error function: 8

Image Statistics This error function is minimized by differentiating with respect to W: Setting the result equal to zero, and solving for w to yield: 9

The log error in the linear predictor is then given by: It is from this error that additional statistics are collected, namely the mean, variance, skewness and kurtosis. 10

Classification Linear Separable SVM Linear Non-Separable SVM 11

Classification Non-Linear SVM 12

Results Use 640*480 pixel, message consists of a n*n pixel n= {32,64,128,256},pixel range [0,255]. 13

Conclusion These higher-order statistics appear to capture certain properties of “natural" images, and more importantly,these statistics are significantly altered when a message is embedded within an image. To avoid detection, of course, one need only embed a small enough message that does not significantly disturb the image statistics. 14

References H. Farid. Detecting hidden messages using higher-order statistical models. In International Conference on Image Processing, page (to appear), Rochester, New York, M. Vetterli. A theory of multirate filter banks. IEEE Tr ansactions on ASSP, 15