Summer 2018 Final Exam Schedule

 

Final Examination Schedule

PLEASE JOIN US AS THE FOLLOWING CANDIDATES PRESENT THEIR CULMINATING WORK.

Summer 2018
 

Tuesday, June 26

Gayathri Palanisami

Chair: Dr. Kelvin Sung
Candidate: Master of Science in Computer Science & Software Engineering
3:00 P.M.; DISC 464
Agglo - A Quantified Self Mobile Application

Quantified Self (QS) is a popular and growing movement that incorporates technology to acquire data of users’ daily activities such as health, fitness, travel, shopping, banking and productivity to provide increased awareness of their daily behavior and habits. A total of 69% of US adults track at least one health metric. Among these adults, almost half are still tracking the statistics in their heads. This capstone project presents a prototype of a full stack mobile-based QS system which aggregates users’ daily activities and provides meaningful visualizations from the raw data. The system supports self-tracking by providing a single source of truth of users’ daily health, fitness, travel and productivity data. The main purpose of this capstone project is to learn and develop a full stack application that provides deeper knowledge on mobile technologies, RESTful APIs, visualization techniques and cloud services. This project takes advantage of iOS technologies, Google Cloud services, RESTful APIs and open source visualization libraries. Through this project, one can learn the underlying technologies involved, different testing methodologies and challenges faced during the design and development of a full stack application. As an evaluation of the learnings, a prototype application in the domain of QS has been developed with the main purposes of self-tracking and multi-source data collection. With the consideration of performance, security, maintainability and usability, the system prototype provides users with a complete set of features such as user login, meaningful visualizations of daily and historical data, and app data management. The results of the preliminary usability testing indicate that the overall user experience and usability of the application are good. 

Thursday, July 12

Bhumikaben Patel

Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering
11:00 A.M.; DISC 464
Prototype System for Handling 3D AR Data Using Spatial Database

The main goal of the project is to design and develop a system that can serve as a database component of Augmented Space Library (ASL). ASL is an ongoing project at the University of Washington Bothell. It supports the augmentation of existing 3D geometries onto communicated scanned world meshes.

To achieve this goal, we developed a prototype application that provides three main functionalities to its users – save objects, delete objects, and find spatially intersecting objects from the database with the user-provided bounding box. To store the real-world 3D input data, into the spatial database, we represent it into the Triangulated Irregular Network (TIN) geometric representation using Well Known Text (WKT) format. TIN is used to represent a continuous surface consisting entirely of triangular facets. To delete the objects from the system, the user provides the input in the form of a 3D mesh file of objects to be removed. To find the intersecting objects, the user enters the coordinates of the input bounding box. The result of spatial intersection is calculated using a spatial operator, which performs the bounding box comparison between the objects. To achieve higher system performance, the GiST-spatial index structure is used. The system provides 100% accuracy for finding intersecting objects. Results are verified in Unity, a cross-platform game engine.

Duncan MacMichael

Chair: Dr. Yang Peng
Candidate: Master of Science in Computer Science & Software Engineering
1:30 P.M.; DISC 464
Implementation Of Ipv6-Based Multihop Communication Over Bluetooth Low Energy On The Universal Windows Platform

As the popularity of the Internet of Things increases, so too does the need for wireless Internet connectivity on a variety of low-powered devices. Two existing technologies are currently positioned to meet this need: Bluetooth Low Energy and IPv6. To join these two technology stacks, several open source operating systems have begun work on connecting them together for IoT devices. However, one of the largest developer platforms available does not presently have this capability: Windows 10 IoT Core. Therefore, the primary goal of this project is to enable multihop transfer of IPv6 network packets over a Bluetooth Low Energy subnet of devices, using Windows 10 and its Universal APIs as the platform. To accomplish this, two software modules and three dynamic link libraries were designed and implemented to bridge the Windows TCP/IP network stack and Bluetooth stack. The two components are a kernel mode Windows Filtering Platform callout driver and a user mode packet processing application. These two components are supported by three libraries: an app-to-driver interoperability library, a Bluetooth Low Energy Generic Attribute library, and an IPv6-Over-Low-power-Personal-Area-Network (6LoWPAN) library. This system was successful in transferring IPv6 data over multiple hops using Bluetooth Low Energy, providing an initial implementation of this solution on Windows 10 that can also serve as a basis for further development on other IoT systems.

Friday, July 20

Kritika V. Kumar 

Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering
1:15 P.M.; DISC 464
Document Summarizer

The rate of rapid information growth has created an urgent need for efficient and accurate summarization systems. By extracting important sentences and creating comprehensive yet concise summaries, it’s possible to quickly assess if a document is worth reading. It is especially useful for students, authors and academic researchers who need to comb through huge numbers of documents every day. This capstone project presents a system which generates summary by using the core natural language processing NLTK , Theano and various text parsing libraries. The system reduces documents in size while retaining their text coherence and meanings. This is accomplished by extracting feature vectors based on term frequency-inverse document frequency (tf-idf), sentence length, sentence position and cohesion from documents, and using a deep learning algorithm called Restricted Boltzmann Machine (RBM) to generate the summary. In addition, the user is given the flexibility to decide the compression ratio of the documents.

Thursday, July 26

Thy Vu

Chair: Dr. Yang Peng
Candidate: Master of Science in Computer Science & Software Engineering
8:45 A.M.; DISC 464
Cross-Region Communication Latency Measurement for Cloud-based IoT Systems

The combination of Internet of Things and cloud services opens up the possibility of large-scale IoT systems that can span across a wide geographical area, some of which are time-sensitive such as intelligent traffic analysis, wearable cognitive assistance, and teleoperation, or drone control. Cloud service resources are also divided by regions, so IoT devices in such system may rely on data sharing from devices managed by different cloud service regions. To assess latency of such wide-spanning IoT system and provide low latency data delivery across service regions, a set of serverless programs are developed in this project to measure the communication latency on delivering IoT data between multiple cloud service regions owned by Amazon Web Service and/or Microsoft Azure. Latency data in regions across United States was collected to analyze the patterns of how region-region latency is attributed across cloud service providers over time. The developed serverless program can also be deployed to edge devices such as Raspberry Pi to enable edge-cloud latency measurement in real-time.

Friday, July 27

Ajay Reddy Palleri Kesavan

Chair: Dr. Brent Lagesse
Candidate: Master of Science in Computer Science & Software Engineering
11:00 A.M.; DISC 464
Randomized Representative Based Election: Defense Against 51% Attacks in Mobile Crowd Sensing Frameworks

Smart devices and wearable have become an epicenter of human lives and have increasingly become more complex and powerful to make people’s life easier. Smart devices like smart phones and wearable like a smart watch today are equipped to provide pervasive connectivity, quality communication and a glut of other services made possible by an array of high-grade sensors like ambient light sensor, proximity sensor, barometer and gyroscope to name a few. This unique coupling between sensor technology and human interaction has a potential to offer a multitude of opportunities and applications in mobile crowd sensing paradigm, such as real-time road traffic monitoring, noise pollution, health monitoring etc. In this paradigm, people become the centerpiece of the sensing process where users can gather data whenever and wherever, using the mobile sensor devices and they own the process of data retrieval and maintaining of the cleanliness of the data. But, humans may be unreliable and malevolent and can hurt the cleanliness of the data being collected for their own benefit, which is why mechanisms for detecting and deterring malevolent activities in mobile crowd sensing become imperative than ever. This paper explore one such attack called 51% attack, where the adversaries take over 51% of the network and influence the consensus of the data collected to obtain incentives. We propose a defence against such an attack by implementing a moving target defense in a Randomized representative based election with a proof of stake payment mechanism to maintain the integrity of the data and reduce monetary loss for the data aggregator during such attacks. To test this method, we simulate an attack by an adversary who gives malicious data and assess their total monetary gain and the effort needed to obtain a profit.

Wednesday, August 8

Aaron Hitchcock

Chair: Dr. Kelvin Sung
Candidate: Master of Science in Computer Science & Software Engineering
3:30 P.M.; DISC 464
Augmented Reality Views Via an Unmanned Aerial Vehicle

The fields of Augmented Reality (AR) and Virtual Reality (VR) have a consistent focus on first person experiences. By utilizing wearable devices like the Microsoft HoloLens, Oculus Rift or Google Cardboard users can enter and interact with a virtual space.  The Augmented Space Library (ASL), developed by Cross Reality Collaboration Sandbox (CRCS) Research group, seeks to combine both physical and virtual spaces with virtual objects and allows multiple users, both local and remote to interact.  The capabilities of both these technologies can be expanded by the use of a remote controlled camera allowing for the addition of third person or remote first person viewing.  This project creates a system and corresponding API allowing for integration of these views into both AR and VR applications as well as the ASL system.  The system gives a user the ability to navigate and explore a remote physical space in real time or see themselves and their surroundings in third person.   From a functional perspective this requires the remote control of a highly maneuverable camera.  This was achieved through the use of a Drone or UAV (Unmanned Aerial Vehicle).   These new virtual views will allow for AR/VR interaction in new ways as all prior physical points of view were bound to the user.

Thursday, August 9

Thomas Matsumi Brown

Chair: Dr. Kelvin Sung
Candidate: Master of Science in Computer Science & Software Engineering
1:15 P.M.; DISC 464
Suggestions for an Improved Approach to the Cross Reality Collaborative Framework

This thesis builds upon the previous iteration of the Cross Reality Collaborative Framework (CRCF) proposed by Doctor Kelvin Sung in early 2017. The work accompanying this thesis analyzes the contributions presented in the original publication for the Cross Reality Collaborative Framework and generates suggestions regarding its potential improvement. A restructuring of the framework to incorporate dimensionality is accompanied by specific definitions of the variable values of state and control that can be encountered for the elements of interaction described. In order to detail and explore the suggestions made in this study, three stages of development are discussed. The first stage is a general overview of the academic corpus involving Mixed Reality applications, frameworks, and technologies. The second stage of development is a restructuring of CRCF utilizing clearer element definitions and explicit state and control boundaries. The final stage involves a qualitative assessment of the changes to CRCF by reviewing user interactions with applications in both an academic and commercial setting. Additionally, several small programs were developed throughout the course of this work that aimed to assist discussion, provide examples for CRCF element state or control values, and provide a demonstration of coding ability. These applications are briefly described to elucidate their functionality and contribution. General trends of CRCF signature designations will follow application interaction review to close the paper and highlight noted areas of improvement for CRCF as a result of the suggestions made in this work.

Tuesday, August 14

Steven Whitham Brown

Chair: Dr. Dong Si
Candidate: Master of Science in Computer Science & Software Engineering
9:00 A.M.; UW1 370
Automated Heartbeat Arrythmia Detection in Zebrafish

According to the World Health Organization, heart disease is the number one cause of death worldwide. Heart related care accounts for the highest financial cost of any health care categories in the United States. Many heart diseases are treatable if detected early, but  many people go untreated because they fail to recognize the early warning signs until a catastrophic episode occurs. Electrocardiograms (ECG) are useful in detecting irregular heartbeats, a symptom of many types of heart disease, but require a patient to be monitored while their heart performs irregular beats. Analysis of these readings requires a doctor to evaluate the data. Automating the detection of irregular heartbeats is a growing research area facing many obstacles. Privacy concerns limit the availability of human heartbeat data to publicly available datasets which provide labeled data used for testing and comparison but do not allow for live analysis of changing heart conditions. To address these limitations researchers at the University of Washington Bothell have been developing a suite of tools to conduct analysis on zebrafish. This project expands on previous work by this team in automating anomaly detection using Convolutional Neural Networks. In this paper, we demonstrate the viability of Recurrent Neural Networks in performing anomaly detection in zebrafish heartbeat patterns and compare the results with previous work.

Back to top