Security enhancement of image steganography using embedded integrity features

Other Title(s)

التحسين الأمني لإسلوب الإخفاء في الصور باستخدام خصائص السلامة المتضمنة

Dissertant

al-Windawi, Zaynah Talat Rashid

Thesis advisor

al-Jarrah, Muzaffar

Comitee Members

Abu Arqub, Abd al-Rahman Muhammad
al-Najdawi, Nijad

University

Middle East University

Faculty

Faculty of Information Technology

Department

Computer Science Department

University Country

Jordan

Degree

Master

Degree Date

2017

English Abstract

Steganography is a security method that hides secret data inside cover media where the very existence of the embedded secret data is not perceptible.

The cover object can be image, audio or video; the most commonly used is an image file.

This thesis presents a model for protecting the security and integrity of secret data embedded in grayscale images, to detect alterations to the secret data that can happen during transmission, and to protect secrecy of the secret data through adding decoy data.

A proposed model is presented in which a secret image is embedded in a grayscale cover image, together with file checksum of the secret data and an extra bit pair per byte to serve integrity verification.

The secret data is read as a stream of bytes and the bytes are split into four pairs of bits, the group of four pairs are hidden inside uncompressed grayscale images using the 4LSB replacement technique.

A decoy bit pair which represents the inverse of the data bit pair is combined with the data bit pair and stored in the right half- byte of a cover byte.

The decoy bit pair serves in verifying that the data bit pair has not been changed, and in protecting the secret data if an adversary manages to uncover the embedded data.

Extraction of the secret file is achieved through merging the four hidden data pairs into bytes.

During the extraction process, a pair comparison between data and decoy pairs is performed to detect alterations.

Also, the file checksum is calculated for the extracted secret data.

A checksum comparison between the embedded and re-calculated checksums is used to detect alterations, as a second detection method, in case an alteration is missed by the pair comparison.

A list of location and content of the altered bytes are produced, to help in investigating the attacks.

This model is implemented in the MATLAB R2015a environment.

The model is evaluated using 1000 BOSSbase public grayscale images.

The secret image size represents 25% of the cover image's size.

The purpose of the evaluation is in two folds, detection accuracy and imperceptibility.

The detection accuracy is the ratio of detected to altered bytes, which was %99.6.

Measuring imperceptibility was based on the Peak-Signal-to-Noise Ratio (PSNR) metric value, between stego and clean images.

The PSNR value was calculated as PSNR1, for embedding straight decoy with the data pair, PSNR2 for embedding with inverted decoy, and PSNR3 for embedding without decoy.

The PSNR3 value was the highest because only 2 bits were embedded.

The PSNR2 value was higher than PSNR1, due to the inversion of the decoy pair.

Minor visual differences were noticed in some clean / stego pairs, due to using grayscale images, however, it will not be observed without close examination of both images.

An alternative 2LSB embedding scheme is proposed where the data and decoy pairs are embedded in alternate bytes, which has eliminated any visual discrepancy

Main Subjects

Information Technology and Computer Science

Topics

• Cryptography

No. of Pages

69

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-762709