51
Thesis Title: Modifid Multi-Level Steganography to Enhance Data Security
Author: Shadi Elshare, Supervisor: Nameer N. EL-Emam, Year: 2018
Faculty: Information Technology, Department: Computer Science

Abstract: Data-hiding using steganography algorithm becomes an important technique to prevent unauthorized users to have access to a secret data. In this thesis, proposed steganography algorithm has been constructed to hide as much as possible from the load of secret data in a color and a gray images, this algorithm is named Deep Hiding Extraction Algorithm (DHEA) to modify Multi-Level Steganography (MLS). This algorithm is based on Modified Least Significant Bit (MDLSB) to scatter data in a cover image. The proposed DHEA algorithm defines a number of levels randomly; where each level using a gray image except the last level that uses a color image. Furthermore, proper randomization approach with two layers have been implemented; the first layer uses random pixels selection for hiding a secret data at each level while the second layer implements at the last level to move randomly from segment to the other. In addition, the proposed approach implements an effective lossless image compression using DEFLATE algorithm to make it possible to hide data into a next level. Dynamic encryption algorithm based on Advanced Encryption Standard (AES) has been applied at each level by changing cipher keys from one level to the next to increase the security and working against attackers. Soft computing with a meta-heuristic approach using Artificial Bee Colony (ABC)algorithm is introduced to achieve smoothing on pixels in image, this approach is effective to reduce the noise caused by a hidden large amount of data and to increase the stego-image quality on the last level. The experimental result demonstrates the effectiveness of the proposed algorithm DHEA and to show high-performing to hide a large amount of data up to 4-bpp (bits per pixel) with high security in terms of hard extraction of a secret message and noise reduction of the stego-image. Moreover, using deep hiding with unlimited levels is promising to confuse attackers and to compress a deep sequence of images into one image.

Keywords: Steganography, Multi-level steganography, Bee Colony Algorithm, least Significant Bit, Image Smoothing, Segmentation Image

52
Thesis Title: System Identification of Quadcopter Using Experimental Data
Author: Iyad Mahmoud Salameh, Supervisor: Dr. Tarek A. Tutunji, Year: 2018
Faculty: Engineering, Department: Mechatronics Engineering

Abstract: An unmanned areal vehicle, Quadcopter, is assembled and tested to develop suitable mathematical models for takeoff and hovering operations. The mathematical dynamic system is highly nonlinear, complex, and inherently unstable. Therefore, System Identification (SysID) methods are used to obtain the mathematical model of the physical system. The models’ outputs are the altitude and the angular velocities (roll, pitch, and yaw). The models’ inputs are the voltage control signals sent to the motors. Two types of models are considered: Single-Input Single-Output (SISO) and Multiple-Input Single-Output (MISO). The input-output data is collected from flight tests, then logged onto MATLAB where the models are developed using two different methods. In the first one, SysID Toolbox in MATLAB is used to develop Auto-Regressive eXogenous (ARX) transfer function. In the second one, Artificial Neural Networks (ANN) are used to construct transfer functions using NN2TF algorithm. The models are compared and validated using different sets of data, and the models with the lowest error are selected. Furthermore, simulation results for the ARX and NN2TF methods are compared, where it is shown that the derived models using NN2TF method have lower orders with small errors.

Keywords: Quadcopter, SysID, SISO, MISO, MATLAB, ARX, NN2TF algorithm

53
Thesis Title: Neural Network Control for Enhanced Response of Thyristor Controlled Reactor Compensator
Author: Dana Mohammed Rafat Ragab, Supervisor: Dr. Jasim Ghaeb, Year: 2018
Faculty: Engineering, Department: Mechatronics Engineering

Abstract: In this work, a Neural Network Control (NNC) is proposed for load voltage balancing in a three-phase electrical power system. The Neural Network (NN) is suggested to determine the appropriate set of firing angles required for the Thyristor Controlled Reactor (TCR) to balance the three load voltages accurately and quickly. In order to validate the performance of the proposed NNC, Aqaba-Qatrana- Amman South (AQAS) power system is considered as a case study and both MATLAB/Simulink and laboratory model are built. Different feeding techniques to irrigate the NN with input data are proposed; RMS values of the three load voltages (RLV), RMS values of the space vector of three load voltages (RSV) and RMS values of both three load voltages and their space vector (RLVSV). These techniques compromise between reducing the measured load parameters and providing qualitative data about system status. Therefore, both the number of required NNs and the complexity of NN structure are reduced significantly. Thus, the response time of the NNC is enhanced and the required firing angles are provided to the TCR in 10 ms for 50 Hz system frequency. It is worth to mention that all calculations associated with feeding techniques are performed in terms of the proposed space vector signal, which has twice of the system frequency. In this work, it is proved that in case of unbalance three phase load change, the variation of space vector takes a sinusoidal form. Furthermore, several simulation and experimental test cases are considered to examine the capability of the NN for generating the required set of firing angles based on Voltage Unbalance Factor (VUF) performance metric. The results show that NN with RLV and RLVSV feeding techniques provides a satisfactory performance in unbalance mitigation compared to the well-established NN techniques.

Keywords: NNC, TCR, RLV, RMS, RSV, RLVSV, VUF

54
Thesis Title: كتاب "أفانين البلاغة" للراغب الأصفهاني - دراسةً وتحقيقًا
Author: عمر ماجد عبد الهادي السنوي, Supervisor: أ.د. محمد حسين عبيد الله, Year: 2018
Faculty: Arts, Department: Arabic Language and Literature

Abstract: تتناول هذه الدراسة كتاب "أفانين البلاغة" للراغب الأصفهاني، دراسة لمضمونه ولمؤلفه، وتحقيقا لمخطوطته، وفق المنهج العلمي الذي سار عليه رواد صنعة تحقيق التراث، وتأتي أهمية هذا الكتاب من علو مكانة المؤلف في هذا الباب، كما تأتي من أهمية الكتاب نفسه، لأنه يعد معلما من معالم علم البلاغة العربية، فقد اشتمل على خلاصة فنون البلاغة في تلك الحقبة، مع سهولة عرضه لهذه الخلاصة من خلال الأمثلة والتعليق عليها، ثم لم يخله من ترجيحاته واختياراته في مسائل عدة.

Keywords: أفانين البلاغة، مخطوطه، تحقيق، بلاغة

55
Thesis Title: From UML/MARTE Component Based Specifications to CSP-OZ Specifications
Author: Khaled Nassir Alsayeh, Supervisor: Mohamed Bettaz, Year: 2018
Faculty: Information Technology, Department: Computer Science

Abstract: UML is a language designed for general purpose modelling. It has profiles that provide generic extension to customize the models for specific languages and domains. MARTE is a UML profile used for modelling Real-time and embedded systems. It has been used to facilitate such systems, also has been utilized to handle intra-concurrency feature that is presented by its component RtUnit (Real-time Unit); is similar to the active object, which may be seen as an autonomous execution resource, able to handle different messages at the same time. However, MARTE does not have the ability to formally verify some properties of the modelled system such as deadlock and liveness. Therefore, this research work proposes the transformation of UML/MARTE models into the CSP-OZ; a formal specification language that combines the Object-Z specification language with process algebra CSP, to solve the problem of formal verification issues. The transformation between UML/MARTE and CSP-OZ will be straightforward, since both of the frameworks are sharing the same core abstraction. This thesis work will contribute in twofold. Firstly, to best of our knowledge this work is the first whose transform UML/MARTE model (RtUnit) to CSP-OZ specification. Secondly, defining the transformation rules from UML/MARTE to CSP-OZ specification. This transformation represents the first step of solving MARTE problem in formal verification issues. As a result, this thesis work presents eight different rules for transforming UML/MARTE component (RtUnit) to CSP-OZ specification, these rules have been applied on the case study (Cruise Control System).

Keywords: UML, MARTE, CSP-OZ