Date of Award

3-2005

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Electrical and Computer Engineering

First Advisor

Richards A. Raines, PhD

Abstract

Steganography is a form of secret communication in which a message is hidden into a harmless cover object, concealing the actual existence of the message. Due to the potential abuse by criminals and terrorists, much research has also gone into the field of steganalysis - the art of detecting and deciphering a hidden message. As many novel steganographic hiding algorithms become publicly known, researchers exploit these methods by finding statistical irregularities between clean digital images and images containing hidden data. This creates an on-going race between the two fields and requires constant countermeasures on the part of steganographers in order to maintain truly covert communication. This research effort extends upon previous work in perturbed quantization (PQ) steganography by examining its applicability to the spatial domain. Several different information-reducing transformations are implemented along with the PQ system to study their effect on the security of the system as well as their effect on the steganographic capacity of the system. Additionally, a new statistical attack is formulated for detecting ± 1 embedding techniques in color images. Results from performing state-of-the-art steganalysis reveal that the system is less detectable than comparable hiding methods. Grayscale images embedded with message payloads of 0.4bpp are detected only 9% more accurately than by random guessing, and color images embedded with payloads of 0.2bpp are successfully detected only 6% more reliably than by random guessing.

AFIT Designator

AFIT-GIA-ENG-05-04

DTIC Accession Number

ADA435207

Share

COinS