Towards Statistically Undetectable Steganography

Abstract

Fundamental asymptotic laws for imperfect steganography were established. The size of the secure pay load was shown to be proportional to the square root of the number of elements (pixels) in the digital media (image); the root rate and the steganographic Fisher information were established as the appropriate measures for evaluating the capacity and for benchmarking steganographic systems in digital media. A general framework for building steganographic systems by minimizing embedding distortion was established by drawing a connection between statistical physics and steganography. The framework, termed the Gibbs construction, allows computing fundamental bounds between distortion and statistical detectability. simulate optimal embedding methods, and construct practical embedding algorithms using syndrome-trellis codes. The framework has essentially narrowed down the design of secure steganographic schemes to the task of designing the distortion function. The PI has also proposed a method for optimizing the distortion function to minimize statistical detectability. Practical merit of the achievements was demonstrated by building: new steganographic methods with markedly improved security w.r.t. existing art for images in spatial and JPEG formats.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jun 30, 2011
Accession Number
ADA567184

Entities

People

  • Jessica Fridrich

Organizations

  • Binghamton University

Tags

Communities of Interest

  • Energy and Power Technologies
  • Human Systems
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Algorithms
  • Coding
  • Computational Science
  • Databases
  • Digital Images
  • Digital Media
  • Distortion
  • False Alarms
  • Information Science
  • Markov Chains
  • Monte Carlo Method
  • Probability Distributions
  • Random Variables
  • Reliability
  • Square Roots
  • Supervised Machine Learning
  • Two Dimensional

Readers

  • Image Processing and Computer Vision.
  • Radio communications and signal processing.
  • Statistical inference.