The student Hala Hassan Mahmoud obtained her Ph.D. degree from the Department of Computer Sciences / University of Technology for her thesis entitled (Universal Image Stego-analytic Techniques for Forensic Applications). The work presented in this thesis is to design a tool for blind image analysis using a deep educational tool called Neural Network (CNN). The frame consists of three main phases: pretreatment that uses the conversion of complex double-wavelet wavelengths with a number of level equal to 5 to give an important advantage and has the ability to emphasize edge detection and suppress detail information. This is an important step to reduce redundancy and data dimensions and finally reduce memory and calculation. The detection accuracy is 0.92857, 0.96875 and 0.98387 when a number of levels equal to 3,4 and 5 are specified respectively for HUGO and UNIWARD algorithms with a load of 0.1 bpp and 0.4 bpp and a lower resolution of 0.90769, 0.92897 and 0.96659 is given for the WOW load algorithm of 0.1 basis points and 0.4 basis points respectively. Level 5 gave the best score compared to other levels for all three algorithms. The second stage was the extraction of features created using the CNN set of five consecutive sets of CNN, and the Leaky Rectified Linear Unit (ReLU) activation function gives the result and problem solving. The last stage was the classification using softmax classified to obtain the last two names are the cover and stego.

This was on Thursday 5/9/2019 at the discussions hall in the department and the discussion committee was composed of (Prof. Abdull Moneim Saleh Rahma and Dr. Iyad Rodhan Abbas) of the University of Technology / Department of Computer Sciences and Prof. Mahmoud Khalil Ibrahim (from Al-Nahrain University / College of Information Engineering), Prof. Qasim Mohammed Hussein of Tikrit University / College of Petroleum and Minerals Engineering and Dr. Alaa Abdul Hamid Abdul Latif from Baghdad University College of Education Ibn Al-Haytham / Department of Computer Science in the presence of the supervisor of the student (Dr. Hanaa Mohsen Ahmed).

Download Abstract