Benchmarking steganographic and steganalysis techniques Electronic Imaging of SPIE 2005...

1

Benchmarking steganographic and steganalysis techniques

Electronic Imaging of SPIE 2005Authors:Kharrazi, Mehdi, Husrev T. Sencar, and Nasir

MemonDepartment of Electrical and Computer Engineering

Polytechnic University, Brooklyn, NY 11201, USA

2

Outline

• Introduction• Universal steganalysis• Embedding techniques• Discussion • References

3

Introduction

• specific steganalysis techniques– good results.– might fail on all other steganographic

algorithms. • Universal steganalysis techniques– less accurately.– acceptable results on new and unseen embedding

algorithms.

4

Introduction

• Receiver operating characteristics (ROC)– organizing classifiers– Visualizing– true positive (TP) hit– true negative (TN) correct rejection– false positive (FP) false alarm– false negative (FN)miss

5

Universal steganalysis

Find and calculate features– BSM(binary similarity measure)– Wavelet Based– Feature Based

Strong classification algorithm

Preprocessing Framing

Feature extraction methods

Classification

6

Embedding techniques

• Outguess• Chooses redundant DCT coefficients which have

minimal effect on the cover image and then embed the message.• Adjust remaining coefficients in order preserve the

original histogram of DCT coefficients.

7

Embedding techniques

• F5• Implementation of matrix encoding, decreases the

necessary number of changes.

8

Embedding techniques

• Model based• Transformed image coefficients into two parts, replaces

the perceptually insignificant component with the coded message signal

9

Embedding techniques

• PQ(Perturbed Quantization Steganography )• Re-compressed with a lower quality factor.• Modified coefficients during compression.

10

Discussion

• Actual embedding rate will be lower do to the embedding overhead incurred when splitting the image into smaller blocks.

• FBS has the best performance among the 3 techniques studied.

11

References

• An introduction to ROC analysis• BSM(binary similarity measure)• Wavelet Based• Feature Based• Outguess• F5• Model based• PQ