CSE Masters and Senior Projects

Data embedding in videos and images has various important applications such as digital rights management (DRM), content authentication, copyright protection, error resiliency and concealment as well as law enforcement. With the high possibility of illegal access and unauthorized content manipulation in shared storage platforms such as cloud data centers and with the risk of encountering different types of attacks during network transmission, videos and other sensitive data are usually transmitted and stored in an encrypted form. Accordingly, the need for data hiding techniques that operate directly on the encrypted video domain has emerged. This work proposes a novel data hiding scheme in encrypted video streams where scrambling and data embedding are performed simultaneously at the encoder side by rotating the motion vectors of the cover video. Then a machine learning solution is proposed at the decoder side to classify the motion vectors to rotated/ unrotated, extract the hidden information bits and reconstruct the original cover video. A sequence-dependent approach is applied where the first part of the video is used for training and model generation. The proposed system is composed of two phases: firstly, the training phase where the model is trained to distinguish between the correctly reconstructed macroblocks and the macroblocks reconstructed using rotated motion vectors. Secondly, the testing phase in which the received motion vectors are rotated using four rotation angles which in turn are used to reconstruct the different candidate macroblocks. At this step, the trained model is applied to identify which of the candidate macroblocks are the ones associated with the true motion vectors. Once the true motion vectors are identified, they are compared to the ones received in the bit stream and thus the embedded bits are extracted, and the video is reconstructed. Experiments are conducted on a number of well-known video sequences after compressing them once with MPEG2 video codec standard and then with HEVC video codec standard. A detailed analysis is provided based on the macroblock type, the number of motion vectors and the type of the encoding sequence. Lastly, the proposed solution is evaluated in terms of classification accuracy, embedding capacity and reconstruction quality.

Digital video forensics refers to the process of analysing, examining, evaluating and comparing a video for use in legal matters and court cases. In digital video forensics, the main aim is to detect and identify video forgery and manipulation to ensure a video’s authenticity and reliability for use in court. This work focuses on passive forensics techniques, namely compression-based digital video forensics. The manipulation of videos can be in the form of frame insertion, deletion, cropping, duplication or frame recompression, which can combine several manipulations. When a video is edited by any of the mentioned techniques, the original encoded bitstream is first decoded, editing is applied and then the video is re-compressed before saving it. This means that by detecting re-compression in videos, we can interpret that the video has undergone some form of manipulation. The least number of recompressions a video can have is double compression, the first results from the device initially capturing the video which compresses it to store it in a suitable format and the second comes from the editing software or tool that re-compresses the video after it has been edited. Such editing can also be done multiple times leading to multiple compressions. Thus, finding out the compression history of a video becomes a very important mean for detecting any manipulation and thereby identifying the trustworthiness and authenticity of a video. Several techniques have been studied and investigated for the accurate classification of double and triple compression in videos based on machine learning and deep learning models with promising results being obtained. In this work, a number of experiments are conducted by using KNN, random forest or bi-LSTM classifiers on a dataset of forged and unforged video sequences. In each of the experiments, performance is evaluated based on the classification accuracy and confusion matrix. Experiments are conducted on MPEG2 and HEVC coded videos using the same re-compression quantization parameter and the results of recompression detection are compared. Experiments are also conducted on HEVC coded videos with the same recompression bitrate and the results obtained are compared to existing solutions in literature. The experimental results revealed that both double compression and triple compression can be accurately detected using the proposed machine learning and deep learning solutions.

Programme for International Student Assessment (PISA) is a standardized test conducted by the OECD to measure 15-year-olds’ abilities in reading, mathematics and science knowledge and skills to meet real-life challenges. A school level analysis of the 2018 PISA test results of the UAE was performed. The raw PISA data was first transformed and cleaned using ETL techniques. This resulted in cleaned data for 259 (6,715 students) schools across the seven Emirates. Unsupervised learning algorithms including k-means, k-medoids, hierarchical clustering and DBSCAN were used to group school into similar clusters. Gradient boosting was then used to determine the key features underlying each clustering. Classification and Regression Tree (CART) was used as a visual explanatory mechanism for each clustering. External validation was determined using Purity, Entropy and Adjusted Rand Index (ARI). Internal validation was carried out using Dunn Index, Silhouette Analysis, GAP Index, Davies-Bouldin’s Index, and Calinski-Harabasz Pseudo F-statistic, and t-distributed Stochastic Neighbor Embedding (t-SNE). Application of the various algorithms resulted in 3 to 6 clusters. According to internal validation metrics, k-means (Calinski-Harabasz Pseudo F-statistic = 122.92) with 6 clusters was the best algorithm followed by K-medoid (Calinski-Harabasz Pseudo F-statistic = 90.55) with 3 clusters. School gender was among the top three most important feature identified across algorithms. School Zones, Council, Urban/Rural Status, or the type of Curriculum did not explain clustering across algorithms with an ARI between 0.01 and 0.13 with one exception. School Gender was the best explanatory mechanism across algorithms with the ARI ranging between 0.48 to 0.92. This suggest that whether a school is female only, male only or mixed was the key explanatory mechanism for clustering schools. Therefore, one recommendation is the exploration of differences between these types of schools in how they yield differing PISA test results. Finally, discrepancies between PISA scores and Ministry of Education’s internal exams were also found in certain clusters and warrant further investigation.

The advent of the Internet of Things (IoT) has brought an unprecedented increase in the number of connected devices. Recently, IoT-based devices have been used in several applications including healthcare, data analytics, smart cities, and many others. Time-sensitive applications, such as Vehicle-to-Vehicle (V2V) communication, led to the need for an Ultra-High Reliable Low Latency Communication (URLLC). Consequently, 5G networks gained massive attention from the research community due to its ability to support enormous amounts of transfer rate. One of the main supporting computing paradigms for IoT is cloud computing, as it offers computing capabilities over the Internet. Nevertheless, cloud computing is unsuitable for time-critical applications. Hence, researchers proposed deploying fog computing as part of the 5G small cells to tackle the deficiencies of cloud computing. Many challenges arise while combining 5G technology and fog computing such as scheduling service requests across small cells to reduce the overall latency. In this work, the scheduling problem is modeled as an optimization problem with the objective of minimizing the overall latency. Furthermore, small cells are decentralized by nature. Therefore, a coordination framework is proposed to handle the interdependency between small cells. Accordingly, the decentralized scheduling problem is mapped to a combinatorial optimization problem. The proposed optimization model is validated using an optimization engine. The scheduling problem is known as an NP-hard problem. Thus, a decentralized heuristic solution is proposed to solve the scheduling problem in polynomial time. The proposed solution integrates a novel Simulated Annealing-Based Scheduling (SABS) and Auction-Based Winner Determination (ABWD) heuristic algorithms. To assess the performance and quality of the proposed heuristic solution, a centralized approach is used as a benchmark. Furthermore, sensitivity analysis is conducted in which the impact of each system parameter on the system behavior is investigated. The results prove the adequacy of the proposed solution as the execution time remained approximately constant, with an average of 726 µs, considering different problem sizes. Moreover, the proposed solution is found to be scalable and accommodates the exponential growth of IoT devices.

The rapid adoption of Internet of Things (IoT) technologies is creating a large number of exposed systems with new security vulnerabilities that may endanger critical assets. This is especially true for applications related to Smart Grid (SG) and energy monitoring. Security issues in the bidirectional exchange of data between edge devices can have severe ramifications if not addressed properly. Edge devices are wireless-enabled microcontrollers typically running embedded operating systems. The resource-constrained nature of edge devices in tandem with IoT network protocols creates many unique security challenges. This work investigates key security issues in IoT systems with special emphasis on edge devices. A generic IoT-based monitoring system based on the Message Queueing Telemetry Transport (MQTT) was designed to simulate the operation of SG and similar IoT monitoring applications. ESP32, ESP8266 and Photon hardware, along with multiple operating systems such as Arduino-core, FreeRTOS, LuaRTOS, TinyOS, MongooseOS and NodeMCU were tested. Severity and impact of vulnerabilities was investigated, and counter measures were proposed. The results are that such edge devices are susceptible to a wide array of attacks such as battery draining, eavesdropping, and data injection even when using Transport Layer Security (TLS). While providing better security, TLS consumed only 1-15% more power than the non-TLS scenarios. However, latency suffered a significant increase of up to 5 folds for TLS/SSL, compared to non-TLS counterparts. Stress testing was conducted with different communication payloads, and frequencies of 1, 10, 100 and 1K bytes, and 0.5, 1 and 2Hz respectively, and yielded similar results.

The Taxi Dispatch problem is a well-known and important problem in the field of transportation and logistics, that has many similarities with other fleet management problems. The objective of the taxi dispatch system is to assign idle taxis to passengers waiting at different geographical locations in a way that maximizes resource utilization while minimizing their operating cost. Traditionally, heuristic rules are used in dispatch problems, mainly because of the simplicity and scalability of the approach. However, at high demand rates, rule-based approaches perform poorly. This encouraged many researchers to build more complex models to tackle the dispatch problem, but most of these models are computationally expensive and cannot scale to handle large fleets. Additionally, most of these approaches are not robust enough for a stochastic environment, which is usually the case with real-world traffic. In this work we model the problem as a Markov Game and solve it using Model-Free Multi-Agent Deep Reinforcement Learning, which is the best approach when the environment is stochastic and there is otherwise no good model for it. The main drawback of reinforcement learning is that it requires too much time and data to learn the optimal policy. In this work we address this issue and strive to improve the efficiency of this algorithm. The curse of dimensionality was broken by representing the state variable as an image which made the complexity independent from the number of taxis and requests and only dependent on the size of the map thus allowing the algorithm to handle large fleets with ease. Using a residual convolutional neural network as Q function approximator allowed the agents to learn complex spatial patterns while seeing only few training samples. We have also found that we can reduce the resolution of the state variable by more than half while losing only 3% of the performance. The proposed algorithm was validated against a rule-based heuristic under different supply-demand ratios, and found to outperform the rule-based technique by a large margin when there is a lack of supply.

Traffic classification is the process of associating network traffic with the application or group of applications that generated it. It is an essential part of network management at datacentres and network operators due to its importance in traffic shaping, bandwidth allocation, and cybersecurity. Several techniques were investigated to classify traffic accurately with methods based on machine learning achieving encouraging results. In this work, we conduct several experiments using naïve Bayes, SVM, KNN, and Random Forest Trees on two traffic datasets, namely UNIBS and UNB, which are both publicly available. While the former dataset was collected in an uncontrolled environment that resembles real network behavior, the latter was captured using a highly controlled environment. In the experiments conducted in this work, we look at the classifiers’ performance and their effect on the classification accuracy and F-score. We also assess the suitability of extracted features using feature selection techniques. Moreover, we determine the optimal percentage of packets within a flow that need to be considered while extracting flow-level features. It is observed that when a larger number of packets is considered, the classification performance improves, but the required processing delay increases. Thus, we argue that 60% of packets in a flow would be a good compromise that ensures high performance in the least possible time. Several graphs are generated during each experiment to investigate the effect of varying each parameter on the classification performance. The results of our experiments indicate that Random Forest outperforms all other algorithms achieving a maximum accuracy of 98.3% and an F-score of 0.93. Finally, since software-based classifiers are usually slow and hence incapable of coping with the increasing amount of traffic within congested networks, we implement a highly pipelined Random Forest classifier on a Field-Programmable Gate Array (FPGA) chip. The implementation makes use of the parallel architecture of the FPGA in accelerating such a time-consuming task. The implemented design is capable of achieving an average throughput of 163.24 Gbps which is more than twice the maximum throughput compared to reported work. This enables datacentres to achieve efficient online traffic classification given the dynamic nature of modern networks.

Abstract to be filled

Abstract to be filled

As Internet of Things (IoT) technology and open source distributed file system applications are evolving, home appliances can be monitored and controlled via an IoT based home gateway from anywhere anytime. These gateways collect energy consumption from home appliances and hence create a large amount of data. Due to the large amount of data being generated, utility companies require platforms that enable them to process, analyze, visualize, and monetize the energy consumption data, and to gain meaningful insights into load profiles. This thesis proposes a smart residential area energy management system that enables home owners and utilities to monitor consumption patterns of each home, community, state, and country. Using open source distributed file system visualization tools, home owners and utilities can monitor their home appliances energy consumption on daily, weekly, monthly, and yearly basis. In addition to this, utilities can also monitor neighborhood community, state, and country wise consumption. The architecture was tested to process data from one million smart meters. This data was synthetically generated based on one year of real consumption data from a home. The big data was stored in a Hadoop cluster. Dimensional modeling was used to develop benchmarking queries to create a real time dashboard consisting of charts, graphs, and reports for home owners and utilities. Both Spark and Hive were used to implement the benchmarking queries and it was found that Spark outperformed Hive. To validate the proposed system outcomes, the results were compared with existing proprietary tools such as IBM’s TimeSeries Informix and relational database management systems.

In recent years, cloud computing has emerged as a practical paradigm for providing IT resources, infrastructure and services. This has led to the establishment of large scale datacenters that have substantial energy demands for their operation. These centers are estimated to have the fastest growing carbon foot print from among all of the information and technology sector. This work investigates optimizing the energy consumption in cloud datacenters by using energy efficient allocation of tasks to resources. The work seeks to develop formal optimization models that minimize the energy consumption of compute resources and evaluates the use of existing optimization solvers in testing these models. Energy consumption of cloud computing datacenters is mainly disbursed by the CPU, memory, disk storage, and network, with the CPU consuming the major portion. Hence, as tasks arrive for processing, these tasks must be scheduled efficiently by the cloud resource allocation mechanism. Here, the scheduling problem is modeled using Integer Linear Programming (ILP) techniques, where models are formulated with the objective of minimizing the total power consumed by the active and idle cores of the servers’ CPUs while meeting a set of constraints. Next, we use these models to carry out a detailed performance comparison between a selected set of Generic ILP and 0-1 Boolean Satisfiability based solvers in solving the ILP formulations. Simulation work is carried out using data centers configured following industry-standard servers specifications. Results indicate the developed models have produced noticeable energy savings when compared to common techniques such as round robin. Furthermore, results also showed that from our selected set of solvers, generic ILP solvers had superior performance when compared to SAT-based ILP solvers especially as the number of tasks and resources grow in size.

Recent availability of very large amounts of educational data in digital format often leads to data overload where it is difficult to determine important trends and patterns beyond those provided by traditional statistical techniques. Therefore, educational data mining (EDM) has emerged. Association mining is a type of EDM technique which is well-known for discovering relationships from data with high scale and velocity, but low variety and veracity. This analysis can be performed at the micro-level (e.g., for teachers), meso-level (e.g., for cohorts of schools), or at macro-levels (e.g., at region, province, or country level). This thesis proposes a methodology for the application of association mining to multi-tier sparse and error-ridden educational data. The methodology uses rule templates and is organized around the four analytical dimensions of people, process, environment, and outcomes. The methodology defines Extract Transform and Load (ETL) processes for this type of data and shows how data from lower levels is aggregated to baskets at higher levels. The proposed methodology was applied to data collected from a large-scale continuous professional development (CPD) process for 2,613 teachers in a developing country. The methodology was used to mine interesting rules which were evaluated using the objective metrics of Support, Confidence, and Lift to determine the quality of rules. The Confidence for each level was set to be at least 0.85. The results are that micro-level analysis (n = 2613 teachers) yielded little or no rules with a very low mean Support of 0.00345 (sd. = 0.00214) and mean Lift 6.98 (sd. = 4.63). The situation remained somewhat the same at the meso-level (n = 1391 schools) with a mean Support of 0.0059 (sd. = 0.00051) and mean Lift of 5.46 (sd. = 3.23). The results were significantly better at the macro level (n = 59 clusters) with a mean Support of 0.089 (sd. = 0.021) and mean Lift of 5.925 (sd. = 2.5). The mined rules discovered several anomalies and fidelity violations in the CPD process at various levels. The methodology was also useful in identifying small groups of teachers (6-8 teachers), schools (8-10 schools), and clusters (4-7 clusters) with common characteristics that can be further administered to help improve the CPD process.

Abstract to be filled

Abstract to be filled

Many methods are proposed for the construction of distinguishing test cases DTCs based on a specification given in the form of a Finite State Machine (FSM). In FSM-based testing, we have a black-box FSM Implementation Under Test (IUT) about which we lack some information, and we want to conclude this information by applying input sequences of DTCs to the IUT, then by observing the corresponding output responses final conclusions about the IUT are drawn. A DTC is adaptive if the next input of a DTC is selected based on the previously observed outputs. In this thesis, we propose an incremental approach for the construction of an adaptive DTC for a given set of states of a nondeterministic FSM. In addition, two heuristics are proposed for the derivation of adaptive DTCs. The first heuristic, called H, uses depth first search for a given fixed height while appropriately utilizing hashing to speed up the search for a DTC. The second heuristic, called Hc, is the same as the first; however, it uses a cost function for ordering the inputs to be considered while conducting the search. Comprehensive experiments are conducted, using both real and randomly generated FSMs, to assess the existence of DTCs and compare the performance of the proposed approaches. A detailed summary of the obtained results is included.
Click here to download this thesis

Down Syndrome (DS), also known as trisomy 21, is a form of cognitive skill deficiency caused by a genetic mutation of the 21st chromosome. The incidence of DS in the UAE has reached 1 in every 319 child births according to the Center for Arab Genomic Studies (CAGS). DS-related challenges include deficiencies in working memory, mental retardation, and the development of Alzheimer’s disease (AD). Assistive technology can facilitate skill assessments undertaken for younger (potentially non-verbal) DS children via Brain Computer Interface (BCI), providing more reliable, quantitatively traceable, rehabilitative, automated and standardized assessment protocols. This project argues that current (modified) versions of standardized assessments and curricula (such as AEPS (Assessment, Evaluation, and Programming System)) , implemented in local centers, do not capture the underlying causes for a) lack of comprehension or b) lack of cognitive skill mastery for DS students. A preliminary protocol for a Down syndrome BCI via electroencephalography (EEG) (as opposed to other methods of brain-imaging techniques such as fMRI) tailored for children, is proposed. EEG caps are used to record variations in neural activity during the administration of cognitive assessments (like the Wisconsin Card Sorting Test [WCST]) – for students over a period of 4 months. A numerical analysis (via Statistical regression modeling, Implicit Function and Squashing Time (IFAST) algorithm and Dynamic Naïve Bayes (DNB)) of correlated EEG signals with gathered scores (Experiment results from AEPS - WCST scores and EEG signals), identified the Inferior Parietal Gyrus (IPG) region to be most influential during the color/shape categorization task. Graphic theoretic analysis revealed EEG patterns mimicking those of socializing communities with their specific functions and manner of evolution in learning processes.

PURPOSE: To build and evaluate an Internet of Things (IoT) architecture that utilizes the Message Queue Telemetry Transport (MQTT) and an end-to-end JavaScript stack with a NoSQL database to process real-time on-board sensor data on smart watches to implement context-aware ubiquitous learning applications. METHODS: A prototype system was designed and implemented to serve curriculum-aligned, real-time assessments utilizing smartwatch sensors that represent a learner’s environmental and bodily context. The system integrates a variety of on-board watch sensors like Global Positioning System (GPS), Pedometer, Light Intensity, Ultra-violet Radiation, and Heart Rate Monitor. Smartwatch program uses JavaScript/HTML5. The JavaScript stack utilizes Node.JS/Express for the middleware and Angular 4 for the teacher administration portal. The system uses Google Classroom as the learning management system, PONTE as the MQTT broker, and CouchDB as the NoSQL database. The performance of the prototype was evaluated on real smart watches under various network conditions (Wi-Fi, 3G, EDGE, and 2G). The backend servers were also assessed for scalability. RESULTS: Without edge analytics, the average worst-case response time of telemetry submission and acknowledgement (160-second interval and about 95 KB sensor data) was an acceptable 4.5 seconds for the 2G Lossy Rural, and 356 milliseconds for Wi-Fi. Under normal conditions, watch CPU utilization was between 30-90% and never exceeded 98% in the worst case. Watch battery depleted on average 8.64% for a half-an-hour ubiquitous learning session. A typical quad-core laptop running broker, middleware, and database had an average CPU utilization rate of 6.25% and the worst case of 25% for serving eight physical watches simultaneously. CONCLUSIONS: The proposed IoT-based architecture for smartwatches seems to be feasible and scalable for context-aware ubiquitous learning applications. However, the system needs to undergo field-testing and further optimization using edge-analytics. Scalability to 100’s or 1000’s of watches should also be investigated because theoretically the architecture should scale.

Abstract to be filled

The goal of ubiquitous learning environments is to move learners out of a classroom and into the real world where learners can engage in experiential and tangible learning. Ubiquitous assessment systems enact student learning in the form of teacher, peer, and system-generated assessments that incorporate physical aspects of objects in outdoor locations. A key component of such systems is a wireless-enabled edge device augmented with various types of sensors to represent the state of physical objects and environments. Most such current systems are built using traditional Internet technologies that are not suited for this purpose, and often lead to cumbersome, unreliable and overly complex designs. This thesis presents a novel generic technical architecture for such systems based on the Internet of Things (IoT) computing paradigm. A commonly used IoT edge device was used to implement four variants of the proposed architecture. The variants were based on Advanced Message Queuing Protocol (AMQP), Constrained Application Protocol (CoAP), Message Queue Telemetry Transport (MQTT), and Extensible Messaging and Presence Protocol (XMPP). Each implementation was evaluated in terms of power consumption, CPU utilization and RAM usage, as well as end-to-end latency and throughput in response to network disturbances. In addition, qualitative aspects of each implementation were analysed based on maximum message size, overhead, security, reliability, and ease of implementation and flexibility. While there were statistical differences in power consumption between the four implementations, the practical difference was negligible. CoAP proved to be the most efficient in terms of CPU and memory utilization but produced the lowest latency in lagged networks only. However, due to payload limitations and lack of reliability features, CoAP was considered ill-suited for most such applications. Among the other three variants, MQTT and AMQP seem more appropriate in terms of qualitative features, although MQTT was more resource efficient in most technical aspects.

Abstract to be filled

In this thesis, the mutants’ elimination problem considered in finite state machine (FSM) based mutation testing, fault diagnosis, and in the assessment of the effectiveness of test suites is targeted. Given a test suite of some test cases usually derived from a specification FSM and a set of mutants (or fault domain), derived from the specification with respect to some assumed types of faults, mutants’ elimination deals with deleting/killing each mutant of the fault domain that has an output behavior different than that of the specification FSM in respect to some test case of the test suite. However, this process is time consuming, especially when the number of considered mutants is huge. Accordingly, three parallel implementations for the considered problem based on the Open Multi-Processing (OpenMP), Message Passing interface (MPI) and the Compute Unified Device Architecture (CUDA) parallel technologies are presented. Comprehensive experiments are conducted to assess the speedup and execution time of the proposed implementations. On average, over all conducted experiments with both randomly generated and real application FSMs, the speedup of OpenMp, MPI, and GPU against sequential implementation equals 6.4, 22.9, and 569.7 times, respectively. The relative speedup of MPI and CUDA with respect to OpenMp equals 3.5 and 121.5 times, respectively; and the relative speedup of CUDA with respect to MPI equals 96.12 times. In addition, the results obtained using real machines are compared with random machines with the same attributes. CUDA implementation is shown to be scalable in terms of considered number of mutants and FSM size. For instance, limited by the used hardware architecture, CUDA easily handled experiments with 500 Million mutants and operated on machines with 9.5 Million transitions. Experiments are also conducted to determine the experimental setup attributes such as test suite length, number of test cases, and attributes related to the parallel implementations such as threads number in OpenMP, processes number in MPI and number of inputs of a test case that will be applied to the mutants in each GPU invocation.
Click here to download this thesis

Data centers, being an integral part of cloud computing infrastructure, deliver huge computational power, storage, reliability, availability, and cost-effective solutions for the cloud applications. A data center network (DCN) topology connects thousands of servers within the datacenter. One of the biggest challenges in DCN is to provide graceful degradation in performance in the event of a link or server failure. This thesis presents two new fault-tolerant DCN topologies derived from the standard Dcell topology. The proposed topologies are cost-effective and scalable as well. The topologies enhance the overall performance of Dcell topology. We also propose a new mechanism to select the optimal path between the hosts using Genetic Algorithm (GA). Performance evaluation of the proposed topologies and techniques is done through a simulation study using realistic intra-datacenter traffic models. The simulation results show that the proposed topologies outperform the standard Dcell topology, resulting in an improvement of at least 5% in throughput even for a small-size network. GA algorithm for the path selection is applied to the two proposed topologies, and it is found that there is a further improvement of about 2% in the throughput of the topologies.

System intelligence can be expressed in terms of how much it helps to make a desired task easy for human beings. The concept can be extended to user interface of computer software which is the cornerstone of human- machine interaction. User interface is the key in determining the user experience with a given system; regardless whether the system is for an industrial control system or a website. It is very much desirable to make the user interface of a software user friendly so that the user can understand and use the software as intuitively as possible. Just as the Machine Intelligence Quotient (MIQ) is used for describing "intelligence" of a system, we would like to devise a systematic scoring system to describe the quality of user interface (UI). In this work, an objective and automated method of evaluating the user interface quality is proposed and ways of comparing against existing subjective heuristic methods are explored. A derivative of Machine Intelligence Quotient (MIQ) method, called User Interface Quotient "UIQ" is used as an objective way of evaluating user interface equality.
Click here to download this thesis

Derivation of input sequences for distinguishing states of a finite state machine (FSM) specification is well studied in the context of FSM-based functional testing. We present three heuristics for the derivation of distinguishing sequences for nondeterministic FSM specifications. The first is based on a cost function that guides the derivation process, and the second is a genetic algorithm that evolves a population of individuals of possible solutions (or input sequences) using a fitness function and a crossover operator specifically tailored for the considered problem. The third heuristic is a mutation based algorithm that considers a candidate distinguishing sequence, and if the candidate is not a distinguishing sequence, then the algorithm tries to find a solution by appropriately mutating the candidate. Experiments are conducted to assess the performance of the proposed heuristics in addition to an existing algorithm, called exact algorithm, that derives distinguishing sequences of optimal length. Performance is assessed with respect to execution time, virtual memory consumption, and quality (length) of obtained sequences. Experiments are conducted using randomly generated machines with various numbers of states, inputs, outputs, and degrees of nondeterminism. Further, we assess the impact of varying the number of states, inputs, outputs, and degree of nondeterminism. Finally, in addition to the three proposed heuristics, we present a parallel multithreaded implementation of the exact algorithm using Open Multi-Processing. Experiments are conducted to assess the performance of the parallel implementation as compared to the sequential using both execution time speedup and efficiency.
Click here to download this thesis

Chronic respiratory diseases are diseases of the airways and other structures of the lung, usually resulting in difficulty in breathing and other symptoms. Chronic obstructive pulmonary disease (COPD) and Asthma are considered to be the most common of respiratory diseases. By taking into consideration the possibility of disease worsening over time and the negative impact on patient’s daily activities, the continuous monitoring and managing of these diseases has become a necessity. Currently, spirometry remains the recommended test for monitoring and diagnosing both, COPD and Asthma. A patient suffering from COPD or Asthma should be able to monitor his disease in order to avoid a worsening condition over time or exacerbation of the disease in severe cases. Proper monitoring requires regular visits to medical centers for spirometry checks, or else the purchase and use of a portable spirometer; both options are costly in terms of money and time. In this work and due to the pervasiveness and advancement of smartphones, we attempt to make use of their built-in sensors and ever increasing computational capabilities to provide patients with a mobile-based spirometer capable of diagnosing and managing COPD and Asthma in a reliable and cost effective manner. We developed a model that allows the computation of two critical lung parameters: FVC and FEV1 by establishing a relationship between the frequency response of human exhalation recorded by mobile microphone, and the actual flow rate. These two parameters and the FEV1/FVC ratio are critical in assessing the progress and status of the diseases. We designed a Pretest Activity that together with these computed lung parameters is used in the diagnosis phase. Sample data used to test the system is collected from patients at both Oriana, and Al Zahra hospitals in Sharjah, United Arab Emirates (UAE), under the supervision of consultant pulmonologists. Results and the medical diagnosis of the implemented system proved to be in very close proximity with those produced by clinical spirometers. Our work is an attempt among many to confirm the notion that mobile Health (m-Health) can and will play an important role within the healthcare industry in the near future.
Click here to download this thesis

Abstract to be filled

In most applications where humans are involved, it is important to augment the interaction between users and the components of these applications. One significant element is the cognitive state of the subjects involved. The cognitive state can be manipulated by the amount of cognitive workload allocated to the working memory. If the assigned cognitive workload is too low, the subject's cognition will be underutilized. In contrast, if the workload is more than the subject's capabilities, he or she will be mentally overloaded. Thus, there is a serious need to accurately assess and quantify cognitive workload levels.In this work, a method for separating four different cognitive workload levels is presented. We use an existing data set that contains EEG signals recorded from sixteen subjects while experiencing four different levels of cognitive workload. Some of these workload levels is due to the degradation of visual stimuli. The proposed solution integrates preprocessing of EEG signals, feature extraction based on discrete wavelet transform and statistical features, dimensionality reduction using stepwise regression and multiclass linear classification. Experimental results show that the average classification accuracy of the presented method is 93.4%. The effect of EEG channel selection on the classification accuracy is also investigated. The results show that channels included in the brain frontal lobes are important in cognitive workload classification. By utilizing only 23 channels, most of them are located in the frontal region; the proposed solution provides an average classification accuracy of 91%. It is shown that the proposed solution is more accurate and computationally less demanding when compared to the existing work.
Click here to download this thesis

Abstract to be filled

Abstract to be filled

Abstract to be filled

Automatic test derivation from formal specifications offers a rigorous discipline to functional conformance testing. In various application domains, such as communication protocols and other reactive systems, the specification can be represented in the form of an Extended Finite State Machine (EFSM). A number of methods can be used for deriving test suites from an EFSM specification. In practice, developing and applying these test suites to an implementation under test is time consuming and costly. Thus, it is desirable to determine high quality test suites in order to reduce the cost of testing. This research aims at determining and comparing the quality of various test suites. Using six realistic application examples, various known types of EFSM based test suites are derived and experiments are conducted to assess the fault coverage of these test suites. The assessment is carried out using EFSM mutants of these specifications, namely, EFSM mutants with single and double transfer faults, single assignment faults and single output parameter faults. The various types of considered test suites include single transfer fault, double transfer fault, all uses, single assignment fault, transition tour, state identifier, edge pair, prime path, prime path with side trip, and random test suites Ranking of the test suites, in terms of fault coverage and in terms of both coverage and test suite length, is established for each considered type of faults.
Click here to download this thesis

Abstract to be filled

Abstract to be filled

Many methods are used for the development of experiments and conformance tests based on the specification given in the form Finite State Machines (FSMs). In FSM-based testing, we have an FSM or a black-box Implementation Under Test (IUT) about which we lack some information, and we want to deduce this information by conducting experiments on the IUT. An experiment consists of applying input sequences, observing corresponding output responses, and drawing conclusions about the IUT. An experiment is adaptive if at each step of the experiment the next input is selected which is based on the previously observed outputs. A distinguishing experiment determines the initial state of the FSM. In this thesis, we consider two implementations of an existing sequential algorithm for deriving the minimal length of an adaptive distinguishing experiment for a nondeterministic FSM. We show that the execution time for both of these implementations grows exponentially as the size or the number of transitions of the FSM increases. Accordingly, in order to obtain a solution in a reasonable time, we develop four parallel implementations of the considered sequential algorithms, namely, a multi-core implementation on Central Processing Unit, two Graphical Processing Unit (GPU) implementations based on the platforms like CUDA and Thrust, respectively, and an implementation on a Network of Workstations (NoWs). Comprehensive experiments are conducted to assess and compare the performance and the speedup of the developed implementations. Based on the results obtained from these experiments, the parallel implementation on a NoW provides the best performance and speedup, followed by the CUDA, then the Thrust, followed by the multi-core CPU implementation.
Click here to download this thesis

Abstract to be filled

Abstract to be filled

Automatic test derivation from formal specifications offers a rigorous discipline to functional conformance testing. In various application domains, such as communication protocols and other reactive systems, the specification can be represented in the form of an Extended Finite State Machine (EFSM). Many methods can be used for deriving test suites from an EFSM specification. In practice, developing and applying these test suites to an Implementation Under Test (IUT) is time consuming and costly. Thus, it is desirable to determine high quality test suites in order to reduce the cost of testing. To this end, in the first part of this thesis, using six realistic application examples, we conduct experiments, assess, determine the fault coverage, and accordingly rank various known types of EFSM-based test suites. While the purpose of conformance testing is to check if an IUT is different from its specification, an interesting, complementary, yet more complex step, is called fault diagnosis or diagnostic testing. The objective of fault diagnosis is to determine the faulty implementation, and thus find the differences between the specification and its implementation. In the second part of this thesis, we present a diagnostic method, conduct experiments, and assess the fault localization capabilities of the EFSM-based test suites considered in the first part of the thesis. The fault localization capability of a test suite is determined for many types of diagnostic candidates, representing possibly faulty EFSM implementations, such as candidates with single or double transfer faults, candidates with single assignment faults, and many other types of candidates. In addition, for each considered test suite, the method determines the diagnostic tests required, in addition to the considered test suite, for locating a faulty EFSM IUT.
Click here to download this thesis

Abstract to be filled

Diabetes is a chronic disease that needs continuous blood glucose monitoring and self-management. The improper control of blood glucose levels in diabetic patients can lead to serious complications such as kidney and heart diseases, strokes, and blindness. The proper treatment of diabetes, on the other hand, can help a person live a long and normal life. On the other hand, tighter glycemic controls increase the risk of developing hypoglycemia, a sudden drop in a patients’ blood glucose levels that can lead to coma and possibly death if proper action is not taken immediately. Continuous Glucose Monitoring (CGM) sensors placed on a patient body measure glucose levels every few minutes. They are also capable of detecting hypoglycemia. Yet detecting hypoglycemia sometimes is too late for a patient to take proper action, so a better approach is predicting the hypoglycemia event before it occurs. Recent research efforts have been made in predicting subcutaneous glucose levels at specific points in the future. Moreover, the models developed used are ill suited for predicting out-of-range glucose values, namely, hypoglycemia and hyperglycemia. Hence, in this research, we use machine learning techniques suitable for predicting hypoglycemia within a prediction horizon of thirty minutes. This period should be long enough to enable the diabetes patients to avoid hypoglycemia by taking proper action. In specific, we use and compare two approaches to perform the hypoglycemia prediction, namely, a time sensitive artificial neural networks (TS-ANN) and tree based temporal classification (TBTC) by applying feature extraction from the patient glucose signal. While the TS-ANN performed reasonably well (with average sensitivity= 80.19%, average specificity= 98.2%, and average accuracy= 97.6%), nevertheless, the TBTC approach outperformed the TS-ANN one with the ability to predict hypoglycemia events accurately (with average sensitivity= 93.9%, average specificity= 98.8, average accuracy= 98.16%) using three aggregate global features; mean, minimum, and difference, and two parameterized event primitives (PEPs), namely the negative slope and local minimum of the glucose signal.
Click here to download this thesis

Object detection is one of the most important tasks in computer vision. It has multiple applications in many different fields such as face detection, video surveillance and traffic sign recognition. Most of these applications are associated with real-time performance constraints. However, the current implementations of object detection algorithms are computationally intensive and far from real-time performance. The problem is further aggravated in an embedded systems environment where most of these applications are deployed. The high computational complexity makes implementing an embedded object detection system with real-time performance a challenging task. Consequently, there is a strong need for dedicated hardware architectures capable of delivering high detection accuracy within an acceptable processing time given the available hardware resources. The presented work investigates the feasibility of implementing an object detection system on a Field Programmable Gate Array (FPGA) platform as a candidate solution for achieving real-time performance in embedded applications. A parallel hardware architecture that accelerates the execution of three algorithms is proposed. The algorithms are: Scale Invariant Feature Transform (SIFT) feature extraction, Bag of Features (BoF) and Support Vector Machine (SVM). The proposed architecture exploits different forms of parallelism inherent in the aforementioned algorithms to reach real-time constraints. A prototype of the proposed architecture is implemented on an FPGA platform and evaluated using two benchmark datasets. On average, the speedup achieved was ×55.06 times when compared with the feature extraction algorithm implemented in pure software. The speedup achieved in the classification algorithm was ×6.64 times. The difference in classification accuracy between our architecture and the software implementation was less than 3%. In comparison to existing hardware solutions, our proposed hardware architecture can detect an additional 380 SFIT features in real-time. Additionally, the hardware resources utilized by our architecture are less than those required by existing solutions.
Click here to download this thesis

Arabic sign language is the most common way of communication between the deaf and the hearing individuals in the Arab world. Due to the lack of knowledge of Arabic sign language among the hearing society, deaf people tend to be isolated. Most of the research in this area is focused on the level of isolated gesture recognition using vision-based or sensor-based approaches. While few recognition systems were proposed for continuous Arabic sign language using vision-based methods, such systems require complex image processing and feature extraction techniques. Therefore, an automatic sensor-based continuous Arabic sign language recognition system is proposed in this thesis in an attempt to facilitate this kind of communication. In order to build this system, we created a dataset of 40 sentences using an 80-word lexicon. It is intended to make this dataset publicly available to the research community. In the dataset, hand movements and gestures are captured using two DG5-VHand data gloves. Next, as part of data labeling in supervised learning, a camera setup was used to synchronize hand gestures with their corresponding words. Having compiled the dataset, low-complexity preprocessing and feature extraction techniques are applied to eliminate the natural temporal dependency of the data. Subsequently, the system model was built using a low-complexity modified k-Nearest Neighbor (KNN) approach. The proposed technique achieved a sentence recognition rate of 98%. Finally, the results were compared in terms of complexity and recognition accuracy against sequential data systems that use common complex methods such as Nonlinear AutoRegressive eXogenous models (NARX) and Hidden Markov Models (HMMs).
Click here to download this thesis

The quality of human-computer interfaces is becoming increasingly important as smart devices are becoming an essential part of our lives. Often what makes or breaks the market success of a device is not the hardware, but the quality and ease-of-use of the user interface of the smart device. Just as it is possible to discuss the intelligence level of machines in terms of their “machine intelligence quotient,” it is becoming increasingly appropriate to discuss the “intelligence level” of a user interface. This new index would provide a quantitative assessment of user interface quality, and would be an indicator for rating the ease-of-use of the human-computer interface. In this study, a framework has been developed for the assessment of “user interface intelligence quotient” and is used to determine the quality of different smartphone interfaces. After conducting 200+ different human-smartphone experiments with popular smartphones and compiling the results using the methodologies developed, the results are compared to the actual opinion of the users. Results indicated that actual user opinions are in line with the calculated “intelligence” value of the smartphones. This study shows that there is a way to develop a “yardstick” to measure user satisfaction by using purely objective parameters. Search Terms: Machine Intelligence Quotient (MIQ), User Intelligence Quotient (UIQ), Mobile, User Interface, Smartphones, Usability, Fuzzy Logic, Sugeno, Mamdani, FIS.
Click here to download this thesis

The smart grid is the integration of the 21st century information and communications technologies with the 20th century traditional power grid. Such integration empowers the electricity utilities and their consumers to play an interactive role to better manage and operate their power consumptions and integrates their renewable energy resources to the grid. As wireless communication is evolving, it is expected that WiMAX will play a major role in the data and commands exchange between generation, transmission, distribution and consumption control and dispatch centers. This thesis proposes the design of two WiMAX network topologies to serve as a wireless communication network for the smart grid. Based on the smart grid applications’ quality of service requirements, network parameters and scheduling, simulation models were developed. The traffic was classified into five priority classes. Three scheduling algorithms namely; class-based weighted fair, class-based deficit weighted round-robin and class-based strict priority scheduling were used to simulate and assess the performance of the proposed models. Simulation results showed that the class-based strict priority queuing is better for the highest priority classes and the class-based weighted fair queuing preserved the quality of service requirements for all classes.
Click here to download this thesis

Social networking services such as Facebook and Twitter and social media hosting websites such as Flickr and YouTube have become increasingly popular in recent years. One key factor to their attractiveness worldwide is that these sites and services allow people to express and share their opinions, likes, and dislikes, freely and openly. The opinions posted range from criticizing politicians to discussing football matches, citing top news, appraising movies, and recommending new products and services such as mobiles, restaurants, and software. This development has fueled a new field known as sentiment analysis and opinion mining with the goal of extracting people’s sentiment from text to assist customers in their purchase decisions and vendors in enhancing their reputation. This emerging field has attracted a large research interest, but most of the existing work focuses on English text. Hence, in this thesis, we studied sentiment analysis of Arabic text retrieved from a well-known social media site, namely Twitter. Specifically, we studied the topic of target-dependent sentiment analysis of Arabic Twitter text, which has not been addressed in Arabic language before. We developed a system that will acquire Arabic text from Twitter and extract users’ opinions towards different topics and products. Key phases of the system are as follows. In the Data Acquisition phase, we collected tweets from Twitter related to specific topics. In the Tweet-Filtering phase, we reduced the noise in the collected tweets data to facilitate the Annotation phase, in which we annotated the collected tweets depending on the specified topic. In the Data Preprocessing phase, we added tags, normalized the words used in tweets, and removed spam tweets. In the Feature identification phase, we extracted stylistic, syntactic, and semantic features, and selected those yielding better results using features selection algorithms. In the Classification phase, the decision to annotate the tweets as negative, positive, or neutral towards a specific topic was made using a trained machine-learning algorithm. Results from different feature sets, classifiers, and datasets are reported in terms of classification accuracy, Kappa statistic, and F-measure.
Click here to download this thesis

This work investigates the feasibility of using the dynamic features of the eyes for biometric identification. Identifying individuals using eye movements is typically limited by a low accuracy, thus preventing this technique from becoming commercially viable. In addition, the human eyes constitute a rich source of information, still only partially understood so far, hence more research is needed to understand exactly what kind of information they can provide, and what technique should be applied to analyze such information. It is also largely unknown what kind of feature will yield accurate data most useful to biometric identification, or which stimuli most influence most the dynamic features of the eyes and their usability as a biometrical trait. We show that, by combining eye movement features and iris constriction and dilation parameters, the dynamic features of the eye can yield a good level of accuracy for biometric systems. The approach consists of recording and categorizing eye movements as well as changes in pupil size into segments consisting of saccades and fixations, and computing for each the many velocity and acceleration features that are used to train the classifier to perform the biometric identification. We tested four types of stimuli to hypothesize which will provide a viable stimulating method for extracting eye features. The results suggest that simple stimuli such as images and graphs can appropriately excite the dynamic features of the eye for the purpose of biometric identification.
Click here to download this thesis

Agile methods have received significant attention in the last ten years and have successfully been applied to many small- to medium-sized projects. They have enjoyed significant popularity amongst developers. Most of the time, the selection of agile methods and practices is based on personal preference or past experience rather than the characteristics of the project at hand. Furthermore, there are no sufficient guidelines for developers to make an appropriate selection. So far, research in this area focuses mainly on specifying the weaknesses and strengths of each method with little analysis of these methods and their practices. It also offers little guidance on how to choose the best-suited practice for a certain project. We believe that finding a way to link project properties and characteristics with the abilities of agile practices is of great importance. In this thesis, we try to find and propose a methodology for developing customized agile approaches by selecting the best agile practices for a given project. We also implement this methodology into an operational model.
Click here to download this thesis

Obstructive sleep apnea (OSA) is a serious sleep disorder which is characterized by frequent obstruction of the upper airway, often resulting in oxygen desaturation. The serious negative impact of OSA on human health makes monitoring and diagnosing it a necessity. Currently, polysomnography is considered the golden standard for diagnosing OSA, which requires an expensive attended overnight stay at a hospital with considerable wiring between the human body and the system. In the proposed research, we implement a reliable, comfortable, inexpensive, and easily available portable device that allows users to apply the OSA test at home without the need for attended overnight tests. The design takes advantage of a smatrphone’s built-in sensors, pervasiveness, computational capabilities, and user-friendly interface to screen OSA. We use three main sensors to extract physiological signals from patients which are (1) an oximeter to measure the oxygen level, (2) a microphone to record the respiratory efforts, and (3) an accelerometer to detect the body’s movements. The collected signals are then analyzed on the phone to deduce if the patient is suffering from OSA. In the proposed system, we have developed an Android application that is able to record and extract the physiological signals from the patients and analyze them solely on the smartphone without the need for any external resources. The smartphone is able to analyze the oximeter and accelerometer reading. Most health applications use smartphones to collect physiological readings, and then off load them to an external server for analysis. However, in this work we developed an integrated environment that collects and processes data on the smartphone, including the signal processing functions that analyze the recorded respiratory efforts. Finally, we examine our system's ability to screen the disease when compared to the golden standard by testing it on 17 samples. The results showed that 100% of patients were correctly identified as having the disease, and 85.7% of patients were correctly identified as not having the disease. These preliminary results demonstrate the effectiveness of the proposed system as compared to the golden standard and emphasize the important role of smartphones in healthcare.
Click here to download this thesis

Battery technology has not advanced rapidly enough to keep pace with the growing energy demands of today‟s portable electronics. Leading this critical need for energy are smartphone devices which are being deployed and adopted at an increasing rate. Developing sound energy management techniques for these devices requires a good understanding of where and how battery energy is being utilized. Power consumption modeling is therefore crucial for understanding the inner working of these devices and for developing energy-efficient software to run on them. In this work, we attempt to develop power models for Android-based smartphones. A logger application is developed to monitor users‟ activity and collect data related to this activity. We then utilize regression techniques and neural networks to develop power models that relate power consumption to usage behavior. We demonstrate the feasibility of applying the two approaches and provide a detailed comparison between them.
Click here to download this thesis

Developing and selecting an appropriate test suite is an important issue for testing implementations of protocols and other reactive software systems. Many methods are known for the derivation of test suites based on a specification given in the form of Extended Finite State Machine (EFSM). In practice, developing test suites and applying these test suites to an implementation under test is time consuming and costly. Thus, determining high quality test suites reduces the cost of software testing. To this end, in this thesis, we first assess and compare the coverage of test suites derived using known EFSM-Based test derivation criteria and test suites derived using the traditional Data-Flow and Control-Flow criteria. In addition, we assess and compare the coverage of these test suites with randomly generated test suites. Finally, we propose an EFSM-Based test derivation method that derives tests with the guaranteed coverage of transfer faults. Experiments comparing the fault detection capability of derived tests with those derived using the considered EFSM-Based, random, and the traditional Data-Flow and Control-Flow testing criteria are presented.
Click here to download this thesis

The ever increasing health care costs are becoming a major concern to both, individuals and authorities. This has tempted researchers to seek alternative models to the traditional and costly hospital-based monitoring and caring approach. One such an approach is the utilization of mobile units that allow for the remote observation and diagnosis of patients in their homes. Advances in VLSI circuits, single-chip embedded-system computing platforms, mobile telecommunications, and web services have provided valuable opportunities to enhance the design and performance of mobile patient‟s health monitoring platforms. In particular, Radio Frequency Identification (RFID) technology has emerged as one of the possible valuable solutions that can be utilized in future healthcare systems. RFID tags integrated with built-in vital signs sensors such as Body Temperature (TEMP), Blood Pressure (BP), Heart Rate (HR), Blood Sugar Level (BS) and Oxygen Saturation in Blood (SPO2) are useful in identifying and recording the state of a patient. In this work, we proposes the design, implementation, and testing of a mobile RFID-based health care system. The system consists of a wireless mobile vital signs data acquisition unit and a fuzzy-logic–based-software algorithm to monitors and assess patients‟ conditions on 24/7 bases. A set of fuzzy rules are developed to diagnose the monitored patient‟s status based on the received vital signs namely; TEMP, BP, HR, BS, and SPO2. The fuzzy algorithm will output an early warning of any patient‟s abnormality status. The system was implemented and tested at Rashid Center for Diabetes and Research (RCDR) hosted in Khalifa Hospital, Ajman, UAE using a representative sample of 26 patients. System performance is compared with the medically accepted standard, namely, the Modified Early Warning System (MEWS) that is currently widely used in practice. The proposed system has proven that it outperforms the MEWS system in many cases, and hence an indication of the usefulness of this fuzzy-based approach.
Click here to download this thesis

In the near future, data rates of 100 Gb/s and 400 Gb/s will be used to match the increase in bandwidth demand for capacity. Wavelength division multiplexing (WDM) systems transmit multiple wavelengths simultaneously at high data rates over long distances where the signal passes through multiple optical add drop multiplexers (OADMs) along the fiber link towards the destination. The transmitted signals su_er from dispersion induced from the fiber and OADMs, where these e_ects are an important limiting factor. The success of high-bit rate, long-haul, point-to-point optical transmission networks depends on the management of the fiber’s linear and non-linear e_ects. In this thesis, we propose to study the impact of cascaded filters as the signals pass through multiple OADMs to determine its e_ect on the next generation network’s data rates. We aim to study the impact of cascaded filters on single-carrier and dual-carrier 100 and 400 Gb/s optical transmission systems. The eye closure penalty (ECP) will be used as a performance evaluation metric. The results indicate that the filter cascade has a severe impact on the performance of dual-carrier systems relative to the case of single-carrier systems. Secondly, chromatic dispersion (CD) e_ect will be mitigated electronically for 100 and 400 Gb/s systems using fiber-dispersion finite impulse response (FD-FIR) filter. The compensating FIR filter’s coe_cient will be computed from the impulse response of the inverse of the fiber’s transfer function. Bit error rate (BER) versus optical signalto- noise ratio (OSNR) curve will be used to evaluate the compensating technique. The results indicate that for 100 Gb/s PM-QPSK systems, using 2 samples/symbol with maximum number of taps is the best approach to compensate for CD. While for 400 Gb/s PM-16QAM systems, using 4 samples/symbol with 50% of the maximum number of taps is the best approach to compensate for CD.
Click here to download this thesis

In recent years, there have been several improvements in the performance of Integer Linear Programming (ILP) and Boolean Satisfiability (SAT) solvers. These improvements have encouraged the modeling of complex engineering problems as ILP problems. These engineering problems are diverse in nature and include genetics, optimization of power consumption, scheduling, cryptography, and more. One such problem is the ‗clustering problem‘ in Mobile Ad-Hoc Networks (MANETs). The clustering problem in MANETs consists of selecting the most suitable nodes of a given MANET topology as clusterheads and ensuring that regular nodes are connected to clusterheads in such a way that the network lifetime is maximized. This thesis focuses on assessing the performance of state-of-the art generic ILP and 0-1 SAT-based ILP solvers in solving ILP formulations of the clustering problem. The thesis consists of four parts. The first part of this thesis consists of improving the existing ILP formulations of the clustering problem. The second part involves enhancing the ILP formulation of the clustering problem through the addition of intra-cluster communication, coverage constraints and multihop links. The third part focuses on the development of an improved tool to enable conversion of user-created on-screen topologies to an ILP formulation. The fourth and final part of this thesis is the detailed performance comparison of a selected set of Generic ILP and 0-1 SAT-based ILP solvers in solving the improved ILP formulations of the clustering problem generated using the tool. The results obtained indicate that from our selected set of solvers, generic ILP solvers are able to handle relatively large scale MANET topologies, while 0-1 SAT-based ILP solvers are the fastest, for small scale networks. For small scale networks the proposed ILP formulations, such as the Star-Ring base model, together with the high performance solvers would be suitable for use in real-world environments. However for large scale networks, as the time to cluster the network grows exponentially, the solvers will be unable to cluster the network in accordance with the demands of a real-world environment.
Click here to download this thesis

In a mobile wireless network, energy saving of mobile devices is one of the most important features for the extension of devices’ life-time and the network. In mobile networks, the device is expected to have several connections, each with different QoS (Quality of Service) requirements. Meeting the QoS requirements on such devices along with better power saving is a challenging task. Moreover, in realtime scenarios, connections are expected to join and leave the network randomly. Before admitting a connection to the network, its QoS requirements must be checked to make sure that the network has adequate resources to accommodate it. Without a proper call admission control mechanism, the system cannot provide the promised QoS to the real-time applications. This research proposes a scheduling algorithm and a call admission control policy for IEEE 802.16e broadband wireless access standard. The proposed scheduling algorithm is designed towards minimizing power consumption at mobile stations, while maintaining different QoS requirements for real-time traffic. The proposed algorithm considers the dynamic nature of connection joining and termination. Connections will be allowed to join the network only if their QoS parameters can be met without violating those of existing connections. Simulation results show that when QoS delay requirements of the connections are not too restrictive, power savings of approximately 75% and 50% at the mobile station can be achieved for low- and moderate- rate Unsolicited Grant Service traffic types, respectively.
Click here to download this thesis

Wireless Sensor Networks (WSNs) have attracted research interest in recent years due to the significance of the field of applications and the advances in sensor technology. In areas where catastrophic events occur such as environmental disasters and battle fields, the network infrastructure is lost and there is an urgent need to build a network in order to monitor the area and to help in rescue operations or troops deployment. An easy and fast way is to scatter scalar and video sensor nodes in an adhoc manner in the area of interest in order to establish a multimedia wireless sensor network (MWSN). Video sensor nodes provide better coverage of the area and enhance the interpretation of the monitored phenomenon. Two main challenges faced in MWSNs are quality of service (QoS) constraints in addition to energy constraints. Many routing protocols with various routing metrics have been developed for WSNs. However, limited research has been done on MWSN routing protocols and there is room for improvement in this area. Moreover, these routing protocols assume structured network architecture where deployment of nodes is pre-planned. Limited research has been done on routing protocol for MWSN deployed in ad-hoc manner that meets QoS requirements and at the same time considers energy efficiency for the purpose of prolonging the lifetime of the network. In this thesis, an optimal routing protocol is developed for MWSN that is energy-aware and QoS-aware. This routing protocol uses ant colony optimization to find the optimal routing path that maximizes the end-to-end path quality and reliability as well as the network lifetime. End-to-end delay is a constraint set depending on the application used. Each metric used in the path cost can be attributed an importance that varies depending on the application requirements. A simulation model is developed to implement the proposed ant colony optimization algorithm in MWSN. A detailed analysis of various parameters used in ant colony optimization is performed. The proposed algorithm is analyzed and its performance is evaluated. The proposed routing protocol not only provides an optimal path in terms of the QoS and energy metrics but it also has the flexibility to be used in various applications by adjusting the weights for the optimal path metrics based on their importance to the application.
Click here to download this thesis