12/07/2021
By Dalila Megherbi

Othmane Habbouli will defend his Ph.D. thesis proposal in Computer Engineering, titled " Analysis, design, and implementation of a new transparent, high-capacity, secure, robust-to-noise digital information hiding and authentication scheme with tampering detection and automatic recovery via DCT moments,” on Wednesday, Dec. 22 at 1 p.m.

Location: This will be a virtual defense via Zoom. Those interested in attending should contact Committee Chair Dalila_Megherbi@uml.edu and Othmane_Habbouli@student.uml.edu at least 24 hours before the defense to request access to the meeting.

Committee Chair: D. B. Megherbi, ECE Department, (CMINDS), UML, (Ph.D. Thesis Advisor).

Committee Members

  • Kanti Prasad, ECE Department, UML
  • Xuejun Lu, ECE Department, UML

Abstract:

The world is moving at a rapid pace to adapt to digitization as a new way of conducting business on a day-to-day basis. Therefore, the need to process and transfer data over the internet using digital media has become a necessity in this time and era. This transfer and sharing of data information such as digital images, streaming videos, and others have exponentially increased over the internet in recent years. With this increase, digital data becomes more vulnerable to security risks as it can be hacked or tampered with before it makes its way to the end receiver. It is almost impossible to avoid unpredicted attacks such as cropping and unwanted added noise to images caused by either digital circuits or by hackers during transmission. Other challenges, which include low capacity and high security of the hidden information, remain unresolved and need new solutions. In particular, scientists and researchers have been working tirelessly to protect critical information that could jeopardize or threaten our nation's national security in many homeland security applications. Two main security techniques growing fast are watermarking and data information hiding. There are usually two kinds of data hiding schemes that are widely used: spatial domain (based on pixel values) and transform domain (based on frequency components). One of the major existing transform domain schemes is the Discrete Cosine Transform (DCT). A key contributor to the DCT transform is its efficiency and compactness that produces a high accuracy of the reconstructed image with a minimum number of coefficients. However, this transform has a major drawback: its high sensitivity to a low level of unpredicted noise.

In this thesis, we propose a new hybrid information hiding scheme based on DCT moments that combines the spatial domain and transform domain in a single algorithm for an optimal result. The novel proposed method proves to solve some existing problems related to security, noise robustness, high capacity, and self-recovery. The key features of the proposed scheme are the high capability of the hidden information of the same size as the carrier image. In order to achieve redundancy, our scheme can take advantage of the DCT compactness and hide eight different gray-scale images in the carrier image with very high imperceptibility. We show that when using color carrier images with the proposed scheme, the capacity of hidden image information increases to twenty-four gray-scale images while maintaining transparency. We also show the effect of attacks like cropping and how our proposed algorithm can self-recover hidden information with high accuracy. The effect of compression and cryptography on the extracted hidden data information is also tackled. We show that the proposed scheme withstands unwanted noise with or without filtering and other signal processing distortions and outperforms existing classical schemes. A more robust mathematical and experimental analysis on DCT moments in terms of their image reconstruction capability, compactness, tampering detection, and noise sensitivity is investigated. We also analyze and compare the proposed scheme's performance using DCT compared to using Chebyshev and Legendre orthogonal polynomials and their moments. Finally, we analyze and compare the effect of noise on the proposed scheme with Chebyshev and Legendre orthogonal polynomials and moments. This involves modeling and mathematical analysis of the effect of noise on the quality of the extracted reconstructed hidden secret image. The results show the potential value of the proposed methodology for robust cybersecurity applications against unauthorized hacker tampering.

All interested students and faculty members are invited to attend the online defense via remote access.