A Unified Steganalysis Framework

Abstract

Steganalysis research is mainly motivated to investigate the possible features that yields highest accuracy in identifying cover objects subjected to steganographic embedding. Although design of better steganalysis techniques is a crucial goal, given the diversity of steganography techniques and their ever-increasing sophistication, it is not realistic to assume a single technique will outperform others in identifying all types of steganographic techniques. Therefore in practical application scenarios, these competing techniques have to be incorporate together. From this perspective, the most important goal of a steganalysis system, that combines many individual steganalysis techniques, is to improve performance. This project focused on real-world deployment of steganalysis systems. Motivated by this problem, we pursue two fundamental questions: First, how to incorporate several steganalyzers together in a steganalysis system, and second, how to limit the testing cost of a steganalyzer.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 01, 2013
Accession Number
ADA583627

Entities

People

  • Nasir Memon

Tags

Communities of Interest

  • Autonomy
  • Human Systems

DTIC Thesaurus Topics

  • Accuracy
  • Acquisition
  • Air Force
  • Air Force Research Laboratories
  • Algorithms
  • Computational Complexity
  • Data Sets
  • Electronic Mail
  • Feature Extraction
  • Feature Selection
  • Information Science
  • Machine Learning
  • Order Statistics
  • Probability
  • Statistics
  • Supervised Machine Learning
  • Training

Fields of Study

  • Computer science
  • Engineering

Readers

  • Neural Network Machine Learning.
  • Radio communications and signal processing.