Links to the recoded presentation and student posters will be posted on this site.
Thursday, April 29, 2021, 11 am
Judging starts. The judges will review posted videos and posters.
Thursday, April 29, 2021, 5:00 pm - 5:30 pm
Introduction of Keynote Speakers (Dean Chastine) Keynote Speakers: Joe Drury, VP, Software Engineering Robert Fortunato, AVP, Infrastructure Products Yola Belisario, Staff DevOps Engineer
Thursday, April 29, 2021, 5:30 pm - 7:00 pm
Judges must attend assigned Q&A sessions. Each session will have five-six projects.
The team will have 10 min to present the project (2-3 min) and answer questions (7-8
min).
Friday, April 30, 2021, 11 am
Deadline for judges to submit completed score sheets.
Tuesday, May 4, 2021
Winners of the awards announced
Best UndergraduateProject
Best Graduate Project
Best Undergraduate Research Project
Best Graduate Research Project
Assurant Award
Sponsors:
Terabyte Kilobyte
Spring 2021 C-Day Projects
Keynote Speakers
Joe Drury, VP, Software EngineeringJoe Drury, VP, Software Engineering:
Joe Drury is from Marietta, GA and graduated from Southern Polytechnic State University
in 1996 with a Bachelor of Science in Computer Science and earned his MBA from KSU
in 2016. He has a passion for creating solutions and programming languages which
include C, Python, C#, Java, and several others. His background includes working
as a developer in the cellular telephone industry at Alltel and Lucent Technologies
and with billing and customer service systems at LHS and Daleen Technologies. Joe
has been with Assurant since 2001 and has extensive experience with call center and
risk management systems. He currently lives in Canton, GA with his wife and children.
Robert Fortunato, AVP, Infrastructure Products:
Robert Fortunato, AVP, Infrastructure ProductsRobert J. Fortunato Jr. is a senior Fortune 500 technologist, organizational leader,
and researching computer scientist with over 20 years of enterprise experience across
the Fortune 500 and public sector higher-ed spaces. Robert currently focuses on enabling
the direct provisioning, use, and management of enterprise technology by its consumers.
Prior to his current tenure, Robert led several teams of engineers in both the operational
break/fix support and strategic solution design of all aspects of Assurant’s global
technology infrastructure. Before that, Robert was the principal architect and manager
of Assurant’s global SQL Server platform. Prior to Assurant, Robert was at Miami
Dade College (MDC), the US’s second largest institution of higher education, and,
the largest college in the state of Florida. At MDC, Robert was a senior member of
a product-oriented research and development team where he wrote, architected, and
managed extensive software systems and infrastructure platforms. Robert possesses
several undergraduate degrees in Computer Science, holds a Master’s Degree in Information
Technology, has deferred conferral of his Master’s Degree in Computer Science given
his postgraduate studies, and is close to completing his Ph.D. in Computer Science.
Yola Belisario, Staff DevOps Engineer:
Yola Belisario, Staff DevOps EngineerYolanda (Yola) Belisario is from Caracas, Venezuela. She graduated from Florida International
University (FIU) in 2019 with a Bachelor of Science in Computer Science and earned
her M.S. in Computer Science from FIU in 2020 focused in AI and Automation. Yola joined
Assurant through the Talent Development Pipeline (TDP) in her senior undergraduate
semester and since then, she has been promoted twice to an IT Resident and Staff DevOps
Engineer Coach where she is able to intersect her two passions; technology and helping
people. Yola’s experience is focused mainly in creating software for automating the
provision of Infrastructure and Cloud Services. From 9-5 she works tirelessly to make
a positive impact in her team, and at night alone with her sister, she manages the
Instagram account @TheCodingSisters where they focus on breaking stigmas in the tech
industry and empower people to pursue their dreams.
Judges & Guests
Assurant Judjes:
Alexander Mohamed - DevSecOps Engineer - Assurant
Amer Uttamchandani - Director of Infrastructure Products - Assurant
Judges: Lenin Disla, Lori Aakre, Andrew Hamilton GC-28 Modern Web Scraping(Graduate Capstone) byKenny Randolph,Joselyn Giron,Denise Tucker,Justin B Bridges,Sandhya Bantu Abstract:This project was developed for the IT7993 Capstone class in the May semester of 2021.The
goal of the project is to scrape all names of key professionals of organizations in
the open990.org website and insert that information into a structured database for
query and analysis. The Key Professionals dataset aims to include global coverage
of key investor and consultant professionals, beginning with US-based companies, involved
in making an investment decision. The overarching aim of this project is to create
a one-stop center for institutional asset management distribution intelligence; the
one spot to go for mandates, documentation and profiles of consultants, investors,
and managers with key technical contact information by including coverage within the
eVestment network for US investors and consultants. From end to end, the key professional
database project consists of creating a web crawler to retrieve information from the
open990 website, wrangling the data into the desired structure, and inserting it into
a database for comprehensive data analysis. The primary data source is the open990.org
website. The team was given a list of names of organizations as targets to scrape
information. Each organization has a page within the open990 website with the organization
information, including names of the key professionals, which is the target data. Scraping
data from the open990 website consisted of several challenges. First, the website
is coded completely using JavaScript which requires specific techniques to render
and scrape. Second, the different organization sites have different data structures,
which causes problems for parsing. Third, most of the data is in tables that are delivered
through a backend API. Fourth, due to delivery of the tables from a backend API, the
HTML tags used for the data are not unique, so that identifying and parsing specific
data using HTML tags was not possible. Lastly, by observing the network traffic using
the Chrome browser tools, and examining the HAR data returned from Splash, we discovered
the website is delivered through Cloudflare servers, which we believe blocked some
of our attempts to scrape the data. Cloudflare is a network for content delivery featuring
robust security services. The complexity of the webpage is an example of how modern,
secure web development will change the landscape and require webscrapers to develop
more advanced methods of automation. Advisors(s): Dr Meng Han Topic(s): Data/Data Analytics IT 7993Department: Information Technology Presentation|Poster|More Information
Judges: Lenin Disla, Lori Aakre, Andrew Hamilton GC-47 Key Professional Dataset - Dataspider(Graduate Capstone) byJanell Westmoreland,Vy Duong,Nyong Nkereuwem,Kajal S Vaghani,Ritu Choudhary Abstract:The purpose of this project is to build a Web Crawler to extract personal information
from a public website like Reddit and LinkedIn. We completed the Instagram crawling
as a bonus for the project. The team will be using MySQL or any other open source
relational database to organize the data and conduct a quantitative data analysis
on it. Advisors(s): Dr. Han - Professor IT7993 Capstone Jing Wang -Project Sponsor Topic(s): Data/Data Analytics IT7993Department: Information Technology Presentation|Poster|More Information
Judges: Lenin Disla. Lori Aakre, Andrew Hamilton GC-52 KSU Spring Capstone - IT Security Solution for Small Business – Group 1(Graduate Capstone) byWilliam Simmons (other members are not registered on CMT3: Olajumoke Giwa, Beau Beard and Collin Peters) Abstract:For this project, we decided to use the nopCommerce open-source eCommerce solution
for our simulated small business. Leveraging the pre-configured web server provide
to our group, we installed the nopCommerce solution package and built our security
program around the web site. Our security program consists of nginx configured web
server and load balancer, SSL certificate for encryption, firewall and a secure backend
SQL server. Advisors(s): Project Sponsor: Dr. Lei Li Course Professor: Dr. Meng Han Topic(s): Security IT 7993Department: Information Technology Presentation|Poster
Judges: Michael Hoefel, Frank Ziller, Jayson Franklin GR-1 Compare Two Off Angle Normalization(Graduate Research) byEmily Ehrlich Abstract:This work investigates different iris normalization techniques to compare their performance
including elliptical normalization and circular normalization after frontal projection
of off-angle iris recognition. Elliptical normalization samples the iris texture using
elliptical segmentation parameters. For circular unwrapping, we first estimate the
gaze deviation using ellipse parameters and the image will be projected back to frontal
view using perspective transformation. Then, we segment the transformed image and
normalize using circular parameters. We further investigate if: (i) elliptical normalization
or circular unwrapping recognition performance is higher, and (ii) the two segmentations
methods in circular unwrapping increase the recognition efficiency. Based on the preliminary
results, the elliptical normalization method shows slightly better recognition performance
in off-angle iris images with 2.2% decrement in the equal error rate. The motivation
of this research is to provide guidance in the construction of a recognition framework
for off-angle iris images with the analysis and comparison of different normalization
methods. MATLAB was used to calculate the hamming distance and accuracy of each normalization
method and to construct the plotted graphs for display. Based on the preliminary results,
the elliptical normalization method shows slightly better recognition performance
in off-angle iris images with 2.2% decrement in the equal error rate. In addition,
perspective projection shifted the distributions of intra and inter class Hamming
distances to left where its average intra-class and inter-class Hamming distance are
0.3070 and 0.4891, respectively compared with 0.3082 and 0.4900 for elliptical normalization. Advisors(s): Dr. Mahmut Karakaya mkarakay@kennesaw.edu Topic(s): Security Department: Computer Science Presentation|Poster
Judges: Michael Hoefel, Frank Ziller, Jayson Franklin GR-29 Wrist Intent Recognition for Stroke Rehabilitation(Graduate Research) bySuman Bharti Abstract:Abstract Hand mentor robotic device is beneficial for stroke patients . This is rehabilitation
technique used in stroke therapy. It strengthens and improves the range of motion
which ultimately improves the quality of life for severely impaired stroke patients.
It is easy to use without assistance and most importantly stroke survivors able to
use independently. Usage of hand mentor device is quite expensive for stroke patients
on hourly basis . Coming up with most efficient deep learning algorithm for sensor
data is motivation to cut down the cost and easy availability usage for stroke patients.
EMG signal is recorded using relevant sensors which provides useful information to
infer muscle movement. In this study, we utilized publicly available EMG signal datasets
recorded from upper limb of human subjects to develop a neural network based model
for the prediction of wrist motion intention. Research Question or Motivation The
Motivation of this study is to train a simple neural network model to accurately predict
three basic wrist motions (extension, flexion and no motion) using optimum number
of EMG sensors. This model can be further deployed to augment the capabilities of
commercially available robotic-assistive rehabilitation devices. Materials and Methods
Sensor-based continuous hand gesture recognition activity requires profound knowledge
about gesture activities from multitudes of low-level sensor readings. There are two
ways to provide the solutions either to go by handcrafted features from sensor data
or use deep learning techniques. The advantage of using deep learning technique is
to utilize the automatic high-level feature extraction with outstanding performance.
However, sensor data requires signal pre-or post-processing such as feature selection,
dimension reduction, denoising, etc. Based on the literature review of many research
papers, we found that 1D Convolutional Neural Network have recently become the state-of-the-art
technique for crucial signal processing applications. 1D CNN is very effective when
we aim to extract features from fixed-length segments of the overall dataset and where
the location of the feature within the segment is not of high relevance. In addition
to this, real-time and low-cost hardware implementation is feasible using 1D CNN.
After a successful literature review on 1D CNN knowing its advantages and benefits
of using over signal. We decided to use 1D CNN on raw EMG signal data. Preliminary
results: Since it is an application-based project, we planned to work in phases to
achieve the long-term goal of benefitting stroke patients using deep learning techniques.
In this initial phase of the study, we utilized publicly available EMG dataset for
hand gestures from UCI Machine Learning Repository to test the performance of the
1D CNN algorithm on gesture classification. We used only 3 labels (hand at rest, wrist
flexion, wrist extension) out of 8 labels in the dataset for our particular application
requirement. This dataset contains 8 EMG channels collected from commercial MYO Thalmic
bracelet device. We first performed an initial analysis to investigate the optimum
number of sensor/channels based on the highest gesture classification accuracy using
KNN, Decision Tree and Naïve Bayes algorithms. As a result of this analysis, we obtained
the optimum channel combination (Ch1, Ch4, Ch5, Ch8) data which generates the best
classification accuracy. We used these 4 sensor datasets to train a 1D CNN with 78/22
train/test split. Dataset contains total 36 subjects. Data with subject number less
than or equal to 28 is considered as training set and data with subject number greater
than 28 is considered as test set. We also performed an optimization study on finding
the optimum time signal window and overlap sizes of 100 ms and 50 ms . We achieved
test accuracy of 97% for the classification accuracy of 3 gestures (hand at rest,
wrist flexion, wrist extension). Advisors(s): Supervisor : Dr. Coskun Tekes Email id : ctekes@kennesaw.edu Topic(s): Artificial Intelligence CS7992Department: Computer Science Presentation|Poster
(not judged) GR-33 Efficient yet Robust Privacy Preservation \\for MPEG-DASH Based Video Streaming(Graduate Research) by Luke A Cranfill Abstract:MPEG-DASH is a video streaming standard that outlines protocols for sending audio
and video content from a server to a client over HTTP. However, it creates an opportunity
for an adversary to invade users' privacy. While a user is watching a video, information
is leaked in the form of meta-data, the size and time that the server sent data to
the user. After a fingerprint of this data is created, the adversary can use this
to identify whether a target user is watching the corresponding video. Only one defense
strategy has been proposed to deal with this problem: differential privacy that adds
sufficient noise in order to muddle the attacks. However, that strategy still suffers
from the trade-off between privacy and efficiency. This paper proposes a novel defense
strategy against the attacks with rigorous privacy and performance goals creating
a private, scalable solution. Our algorithm, No Data are Alone (NDA), is highly efficient.
The experimental results show that our scheme is more than two times as efficient
in terms of excess downloaded video (represented as waste) than the most efficient
differential privacy-based scheme. Additionally, no classifier can achieve an accuracy
above 7.07% against videos obfuscated with our scheme. Advisors(s): Dr. Junggab Son Topic(s): Security Department: Computer Science Presentation|Poster
Judges: Fernando Pujol, Jaspal Sagoo GR-40 Design and Implementation of a Microservices Web-based Architecture for Code
Deployment and Testing(Graduate Research) by Soin Abdoul Kassif Baba M Traore Abstract:Many tech stars like Netflix, Amazon, PayPal, eBay, and Twitter are evolving from
monolithic to a microservice architecture due to the benefits for Agile and DevOps
teams. Microservices architecture can be applied to multiple industries, like IoT,
using containerization. Virtual containers give an ideal environment for developing
and testing IoT technologies. Since the IoT industry has exponential growth, it is
the responsibility of universities to teach IoT with hands-on labs to minimize the
gap between what the students learn and what is on-demand in the job market. That
can be done by using containerization. There are many approaches in the containerization
field, but they can be difficult to use without depth knowledge in virtualization
and code encapsulation. After a deep analysis of the containerization challenges,
we came with an idea of a microservice infrastructure based on Docker, which is an
open- platform for developing, testing, and running applications using containers,
to solve the virtualization and code-encapsulation problem. Our infrastructure will
provide a code development and testing web-based platform that allows users to securely
go in the process of containerization without spending research time in learning virtualization.
So, students and researchers can focus more on the development and testing of algorithms
and codes. For example, it will be easy to develop containers that allow sensors to
connect to an external server in few cliques, or to run a python code in a total isolate
process in minutes without downloading any containerization software. Advisors(s): Dr. Maria Valero mvalero2@kennesaw.edu Dr Hossain Shahriar hshahria@kennesaw.edu Topic(s): IoT/Cloud/Networking Department: Information Technology Presentation|Poster
Judges: Fernando Pujol, Jaspal Sagoo GR-44 An efficient intrusion detection framework based on federated learning for IoT
networks.(Graduate Research) byOsama Shahid Abstract:There are abundant number of IoT devices that are connected on over multiple networks.
These devices can be exposed to multiple different types of network threats. Though,
these devices do have security and software that does act as a wall of protection
we purpose a Federated Learning (FL) approach that would allow detection of threats
of a network for IoT devices. Federated Learning can be best described as decentralized
training. Adhering to the GDPR rules that prevent data from being distributed, FL
addressed the challenge by bringing the ML model to the data rather than the traditional
method where the data had to be extrapolated and taken to the ML model. This type
of setting is ideal for IoT devices (Client) that are connected to the network and
can download the FL model that would allow them to keep their devices more secure.
FL is different from on-device training. Once the Client(s) download the model and
train on their individual local data, the updated model is shared on a central server.
The central server as all the individual models shared by each client on the federated
network. The central server aggregates all these models as one new global model. For
our project we believe this is an ideal setting for IoT devices that are susceptible
to network threats. In our Federated Network we have four clients, and we create a
FL framework that allows each client to train a FL model on their local data. The
model(s) are then aggregated to create a new global model. It is worth noting that
each client has its’ own distinct type of threat. So, when all the models are aggregated,
they have the knowledge of each of these individual models and this new model is capable
to testing and detecting threats that are posed across all the devices. Advisors(s): Dr. Seyedamin Pouriyeh Topic(s): Security Department: Information Technology Presentation|Poster
Judges: Fernando Pujol, Jaspal Sagoo GR-64 How could IoT assist healthcare system during COVID-19 or future pandemics?(Graduate Research) byMohammad Nasajpour Esfahani Abstract:The Internet of Things (IoT), a technology built upon sensors and devices, has shown
great applicability among various domains, especially healthcare. This pandemic has
critically impacted all parts of society including people, health centers, businesses,
authorities, etc. Researchers are attempting to adopt different technologies to mitigate
this virus faster and save more lives. Regarding the great benefits that Internet
of Things (IoT) has brought into different areas within the healthcare domain, this
technology has been performing several main tasks including diagnosing, monitoring,
tracing, disinfecting, and vaccinating to combat this virus. Our research is conducted
of the possible IoT solutions to mitigate the COVID-19 or even future pandemics. We
have demonstrated the applicability of IoT technologies in three main COVID-19 phases
including “early diagnosis, quarantine time, and after recovery.” Along with such
applications, we also review the proposed IoT applications for the main tasks of IoT,
which could be exponentially helpful for fighting against this virus. Advisors(s): Prof. Seyedamin Pouriyeh: supervisor Prof. Reza Meimandi Parizi: second
supervisor Topic(s): IoT/Cloud/Networking Department: Information Technology Presentation|Poster|More Information
Judges: Joel Hagans, Charles Chidi Igwilo GR-34 Defensive Neural Network(Graduate Research) byHongkyu Lee Abstract:Machine learning (ML) algorithms require a massive amount of data. Firms such as Google
and Facebook exploit user's data to deliver a more precise ML-based service. However,
collecting users' data is a risky action because their private data can be leaked
through the transmission. As a remedy, federated learning is introduced. In federated
learning, a central server distributes a machine learning model to users. Each user
trains the model to its data, and send the model back. Later the models are aggregated
and distributed again. Federated learning is more secure in that it emancipates users
from the risk of sending private data directly. Recently, several researchers have
identified that federated learning is vulnerable to inference attacks. The inference
attack is an adversarial algorithm that identifies the training data only by inspecting
an ML model. A successful attack will allow an attacker to know the private data of
users. We proposed defensive federated learning, the federated learning that deters
inference attack. The defensive federated learning hardens the inference attack and
obfuscates original private data into an unrecognizable form to human eyes. Thus,
the success rate of the inference attack decreases, and even if the attack is successful,
what the attacker can see is distorted data that is not decipherable. What important
is, even if the proposed scheme distorts the original data, it still learns from the
distorted data and achieves high classification accuracy. We showed that our proposed
scheme achieved higher model performance and stronger toleration than differential
privacy, which is the only solution for the inference attack. Advisors(s): Dr. Junggab Son Topic(s): Other (explain in the comments section) N/ADepartment: Computer Science Presentation|Poster
Judges: Joel Hagans, Charles Chidi Igwilo GR-38 Energy Cost and Efficiency on Edge Computing: Challenges and Vision(Graduate Research) byKousalya Banka Abstract:The Internet of Things (IoT) has been the key for many advancements in next-generation
technologies for the past few years. With a conceptual grouping of ecosystem elements
such as sensors, actuators, and smart objects connected to perform complex operations
to perform environmental monitoring, intelligent transport system, smart building,
smart cities, and endless other possibilities. Edge computing helps the IoT’s reach
even further and be more robust by connecting multiple censored devices through the
internet and forming powerful computational capabilities. Unfortunately, this computation
level comes at a cost as the devices are constantly being used to communicate and
perform specific actions. Energy efficiency has focused on finding the optimal way
to utilize the latest technologies while retaining the battery power’s longevity.
In this paper, we present an outline of the difficulties engaged with planning energy-efficient
IoT edge devices and depict recent research that has proposed promising answers that
address these challenges. First, we analyze the challenges that IoT devices bring
in terms of energy consumption. Next, we discuss the different approaches such as
computation offloading, modifying the IoT devices’ designs, and the number of algorithms
that help reduce energy consumption and few latest technologies. Finally, we will
look at the case study that outlines the energy-saving techniques in smart grids,
smart cities, electric vehicles, smart home devices, and VR/AR in real time to apply
the concepts proposed. Advisors(s): Dr. Kun Suo Topic(s): IoT/Cloud/Networking Department: Computer Science Presentation|Poster
Judges: Joel Hagans, Charles Chidi Igwilo GR-45 Framework for Collecting Data from specialized IoT devices.(Graduate Research) byMD SAIFUL ISLAM Abstract:The Internet of Things (IoT) is the most significant and blooming technology in the
21st century. IoT has rapidly developed by covering hundreds of applications in the
civil, health, military, and agriculture areas. IoT is based on the collection of
sensor data through an embedded system, and this embedded system uploads the data
on the internet. Devices and sensor technologies connected over a network can monitor
and measure data in real-time. The main challenge is to collect data from IoT devices,
transmit them to store in the Cloud, and later retrieve them at any time for visualization
and data analysis. All these phases need to be secure by following security protocol
to ensure data integrity. In this paper, we present the design of a lightweight and
easy-to-use data collection framework for IoT devices. This framework consists of
collecting data from sensors and sending them to Cloud storage securely and in real-time
for further processing and visualization. Our main objective is to make a data-collecting
platform that will be plug-and-play and secure so that any organization or research
team can use it to collect data from any IoT device for further data analysis. This
framework is expected to help with the data collection from a variety of different
IoT devices. Advisors(s): Dr. Maria Valero, Dr. Hossain Shahriar Topic(s): IoT/Cloud/Networking Department: Information Technology Presentation|Poster
Judges: Cheryl Coleman, Veanne Smith GR-50 Predicting Users' Engagement During Interviews with Biofeedback, Voice, and
Supervised Machine Learning(Graduate Research) by Thaide Huichapa Abstract:Studies show that the quality of the information collected during an elicitation interview,
and consequently the quality of the software product that needs to be developed, highly
depends on the interviewee's engagement. Because of social expectations, interviewees
tend to hide if they are bored or not engaged. To overcome this problem and support
the analyst during the interviews, this research uses biometric data and voice features,
together with supervised machine learning algorithms, to predict the interviewee's
engagement. We built our solution on an experiment consisted of interviewing 31 participants.
We collected the data using an Empatica wristband and a default recorder from a laptop.
After preprocessing the data and segmented them into single question/answer segments,
we used part of them to train different supervised machine learning algorithms and
the remaining to test the algorithms in order to evaluate their effectiveness and
select the one that performs better. Our results show that both biofeedback and voice,
considered individually, and machine learning can be successfully used to predict
participants' engagement. The results of our work, in addition to being used to help
the analyst conduct better interviews by steering the interview based on the participant's
engagement, can also be used to prioritized requirements. Advisors(s): Dr. Paola Spoletini Topic(s): Software Engineering Department: Software Engineering and Game Design and Development Presentation|Poster
Judges: Cheryl Coleman, Veanne Smith GR-53 An Investigation on Non-Invasive Brain-Computer Interfaces: Emotiv Epoc+ Neuroheadset
and Its Effectiveness(Graduate Research) byMd Jobair Hossain Faruk Abstract:Neurotechnology describes as one of the focal points of today’s research around the
domain of Brain-Computer Interfaces (BCI). The primary attempts of BCI research are
to decoding human speech from brain signals and controlling neuro-psychological patterns
that would benefit people suffering from neurological disorders. In this study, we
illustrate the progress of BCI research and present scores of unveiled contemporary
approaches. First, we explore a decoding natural speech approach that is designed
to decode human speech directly from the human brain onto a digital screen introduced
by Facebook Reality Lab and University of California San Francisco. Then, we study
a recently presented visionary project to control the human brain using Brain-Machine
Interfaces (BMI) approach. We also investigate well-known electroencephalography (EEG)
based Emotiv Epoc+ Neuroheadset and present experimental studies to identify six emotional
parameters using brain signals by experimenting the neuroheadset among three human
subjects. Advisors(s): Prof. Maria Valero Prof. Hossain Shahriar Topic(s): Other (explain in the comments section) Department: Information Technology Presentation|Poster|More Information
Judges: Cheryl Coleman, Veanne Smith GC-54 COVID-19 Mortality Prediction using Machine Learning Techniques(Graduate Capstone) by Lindsay Schirato, Kennedy Makina, Dwayne Flanders Abstract:In late 2019, SARS-CoV2 also known as COVID-19 was first identified in the city of
Wuhan, China. This virus can infect a person and without showing any signs of sickness,
can spread of COVID-19 unknowingly. The World Health Organization declared it a global
pandemic in March 2020 because of its far-reaching effects in every part of the world.
Scientists have been working to leverage technology to prevent spread, detection and
vaccine development. With machine learning, models can predict which patient will
most likely have a higher mortality rate. Using WEKA, a machine learning tool and
a data set based on 95,000 Mexican patients with 20 clinical features, our research
applies models to determine which has the most accuracy. Advisors(s): Dr. Seyedamin Pouriyeh Topic(s): Other (explain in the comments section) Machine LearningDepartment: Information Technology Presentation|Poster|More Information
Judges: Rosie Belisario, Juan Huaca GR-23 Machine Learning Techniques for Malware Network Traffic Detection(Graduate Research) by Jermaine Cameron Abstract:Persistent malware variants are a constant threat to computing infrastructure across
all regions and business sectors. Traditional detection systems focus primarily on
signature-based analysis but this approach cannot adequately keep pace with the velocity
and volume of new malware variants that are continuously deployed onto the internet.
Most network traffic detection techniques are focused on analyzing raw packets and
have not deterred the surge of persistent malware. Therefore, it is important to develop
new research techniques that are focused on optimized metadata from malware network
traffic to effectively identify an ever-increasing expanse of malicious software.
Recent research efforts by Letteri et al. have produced a quality data set (MTA-KDD’19)
that is utilized for this research project. New information in the area of malware
network traffic detection is pursued through this research proposal. Specifically,
I seek to find a defensible answer to the following question: Can machine learning
techniques produce highly accurate classification models for malicious network traffic
detection based on analysis of a statistically optimized data set? I believe that
an affirmative answer to this research question provides a beneficial contribution
to the academic community. The principal tool utilized to analyze the optimized data
set for this research project is the Waikato Environment for Knowledge Analysis (WEKA).
There are 64,550 instances and 33 features in the MTA-KDD’19 data set that are analyzed
along with cross-validation and percentage split alternatives. The classification
experiment performed by the authors of the MTA-KDD’19 data set is used as a baseline.
The following machine learning classification models have been applied for this research
investigation: Multilayer Perceptron, Decision Tree, Support Vector Machine, and K-Nearest
Neighbors. The preliminary settings for these machine learning models include 10-fold
cross-validation and 80% train 20% test data split. The Decision Tree classifier produced
the best preliminary result with 100% accuracy when set to run an 80% training 20%
test split and 99.9954% accuracy when set to run 10-fold cross-validation. This preliminary
result has outperformed the results observed in the experiment presented by the authors
of the MTA-KDD’19 data set. Other preliminary metrics illustrate that the selected
models exhibit consistent and highly accurate performance. The multilayer perceptron
classifier produced a preliminary result of 99.3649% accuracy when set to run an 80%
training 20% test split and 99.3416% accuracy when set to run 10-fold cross-validation.
The K-Nearest Neighbor classifier (K=1) produced a preliminary result of 98.9311%
accuracy when set to run an 80% training 20% test split and 99.0024% accuracy when
set to run 10-fold cross-validation. The Support Vector Machine classifier produced
a preliminary result of 97.8081% accuracy when set to run an 80% training 20% test
split and 97.7755% accuracy when set to run 10-fold cross-validation. The final stage
of this research project will include implementation of additional machine learning
methodologies. These methods will include feature selection techniques and ensemble
learning models. Advisors(s): Dr. Seyedamin Pouriyeh Topic(s): Security CYBR 7240Department: Information Technology Presentation|Poster
Judges: Rosie Belisario, Juan Huaca GR-67 Representation Learning for Motion Sequence(Graduate Research) by Saisangararamaleengam Alagapan (other students are not registerd
on CMT3) Abstract:This research project proposes a new deep learning architecture that is used to align
human poses to be used in an exercise assistant system. In short, the assistant system
takes a video feed of a user doing exercise, then provides visual feedback by comparing
the user’s current pose to a professional trainer’s pose that is stored in the system.
We design a new deep architecture to accomplish this task and show better accuracy
and efficiency. Advisors(s): Project Sponsors - Dr. Ying Xie & Dr. Linh Le Project Advisor- Dr. Meng
Han Topic(s): Data/Data Analytics IT 7993Department: Information Technology Presentation|Poster
Judges: Rosie Belisario, Juan Huaca GR-70 Defending data reconstruction through adaptive image augmentation(Graduate Research) bySeunghyeon Shin Abstract:In this paper, we introduce a data augmentation-based defense strategy for preventing
the reconstruction of training data through the exploitation of stolen model gradient.
The collection of training data to a centralized server has been required for the
training of neural networks in traditional machine learning. However, as privacy becomes
a significant concern, the concept of Federated learning is introduced. In federated
learning, a centralized server shares the well-trained neural network and participating
end-users send the gradient back to the server after training without sharing the
sensitive data itself. As the concept of federated learning does not share the original
data that might include sensitive information, it is believed to be safe against privacy
threats. However, several types of research showed that sharing gradient is not safe
for privacy as the data can be reconstructed from the shared gradient. Model inversion
is an exemplary threat against privacy in deep learning that reconstructs training
data from model parameters. Differential privacy is known as a way to prevent stealing
gradient for this type of attack in machine learning, however, adding noise in the
optimization process to preserve privacy generates significant accuracy loss, so balancing
the privacy and utility is required. Our proposed method provides better performance
than the traditional differentially private classification method through the usage
of grid search that finds the optimized augmentation scheme for each data class. In
our research, We found the best augmentations for each class of CIFAR-10 that guarantees
similar or better accuracy exists compared to differentially private stochastic gradient
descent optimization in deep learning. Our research provides model accuracy and attack
accuracy for comparison, which indicates the accuracy of an augmented dataset and
the dataset consists of recovered images with augmentation applied. We aimed to secure
the higher model accuracy and lower attack accuracy than differentially private classification
results. For example, airplane class in CIFAR-10 dataset has 62.33% of model accuracy
and 34.67% of attack accuracy, and it is better than DPSGD results with 56.78% model
accuracy and 44.73% attack accuracy with sigma=0.5 and l2_clip_norm=1.0. Our research
guarantees a better balance between privacy and utility and also show that adaptive
augmentation can be used in various type of dataset in further researches. Advisors(s): Dr. Junggab Son Topic(s): Security Department: Computer Science Presentation|Poster
Judges: Alexander Mohamed, Trevor Sands UC-7 Software Engineer – Clarity LLC(Undergraduate Capstone)) byAmy Mullins Abstract:Clarity makes an app called CaptionMate that does closed captions for phone calls..
During the internship, a website was made to visualize metrics that are collected
on users such as calls made, minutes used, time active, region, age, theme, font,
and platform used. Bar charts are used to show minutes and calls used on days of the
week. 100% bar charts are used to show how much a day contributes to the usage of
the app; a user contributes to minutes, calls and platform usage; and show calls incoming
vs. outgoing. Line graphs were created to show growth in app usage and number of new
users over time. All the data iis displayed in tables too. Who used the CaptionMate
app, how much and when? Where are these users located? what age group do they fall
into? what themes and fonts do they use? What platforms are they on? Visual studio
with C# .NET Core, SQL Server, and ChartJS were used in the making of the website.
C# was used for the backend of the website; it executed queries and stored results
in lists. SQL Server was used to write queries. These results are displayed using
HTML tables. ChartJS was then used to make visualizations after retreiving necessary
data form the HTML tables. Many results were found. Users use more minutes and make
more calls during the week then on weekends. Users use IOS more than any other platform.
They like the gray theme most out of all themes and are located all around the US.
The number of users has been growing slowly since the app was released. Some users
stop using the app after some time of using it while other users do not try to make
calls or make only unsuccessful calls. Advisors(s): Prof. Dawn Tatum dtatum7@kennesaw.edu Topic(s): Data/Data Analytics CSE 4983Department: Computer Science Presentation|Poster
Judges: Alexander Mohamed, Trevor Sands UC-14 TeleClinic(Undergraduate Capstone)) byJay P Bhatt,Lucius Burch,Zekai Fei,Bijoy Shah,Hao Zhang Abstract:TeleClinic is a telemedicine web application that provides ease of access and a medium
for interaction between patients and their respective doctors and administrators.
In particular, this web portal includes a chat feature, an area for medical reports,
an area for appointment requests, and an area for video recordings. Additionally,
TeleClinic meets the requirements prescribed by the Health Insurance Portability and
Accountability Act (H.I.P.A.A.) via upholding data privacy and safeguarding medical
information. To maximize its overall utility, TeleClinic utilizes the React and React
Redux libraries for its front-end and a NoSQL database in Google Firebase for its
back-end. Advisors(s): Dr. Ken Hoganson Topic(s): IoT/Cloud/Networking CS 4850Department: Computer Science Presentation|Poster|More Information
Judges: Alexander Mohamed, Trevor Sands UC-15 Malware Analysis Using Reverse Engineering(Undergraduate Capstone)) byShamour Jones,Cynthia S Marcellus,Andy Pham,Nathan Rowe,Joshua Rowland Abstract:The motivation for this project is driven by evaluation of the different tools on
the market that allow for breaking down executables or binary files, and understanding
what the malware is doing. By reverse-engineering the malware, we can understand its
impact and how to protect against it. Our focus is to understand where different tools
are stronger than others, as well as understand the evolving landscape of malware
and security overall. For this capstone project, we utilized two different tools and
many sample malware files. The methods used to debug the malware are detailed in our
milestone two report and will be expanded upon in our final presentation. At this
point, we've found the tool WinDbg to be the most versatile for binary and executable
debugging. We also evaluated IDA Pro, and understand the many ways in which its graphical
display of data and relationships, equips a researcher with the necessary tools and
information to walk through an executable. Our focus in milestone 3 is to expand our
documentation and guide on malware debugging to the point that it provides a user
the full breadth of information and steps needed to start from scratch and end with
a broken apart piece of malware. We provided much of this as part of the milestone
2 presentation and report, but we will continue to build on it so it's a useful how-to
guide for anyone trying to debug a piece of malicious code. Advisors(s): Dr. Ying Xie yxie2@kennesaw.edu and Dr. Hossain Shahriar hsahria@kennesaw.edu Topic(s): Security IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Yingying Kang, Tammy Schopf UC-6 Covid-19 Data Analysis - Regression(Undergraduate Capstone)) byNoah Druss Abstract:Covid-19 has been arguably the most impactful event in the past century. SARS-Cov-2
is a viral respiratory illness discovered in late 2019 that has spread to almost every
country in the world. It has directly or indirectly affected just about everybody
in the world greatly, causing over 117 million cases and 2.59 million deaths as of
March 2021. This project has focused on the use of different types of linear regression
to both analyze and predict Covid-19 infection data based on different features. First,
simple linear regression was used to predict total deaths based on infections both
globally and by country. Globally, the R^2$ score was .946, while depending on the
country, the R^2$ score .984 which shows a very effective line of best fit. Second,
polynomial regression (with a degree of 3) was used to predict total deaths based
on total infections by country. This was much more effective, with R^2$ scores up
to .9998. Finally, multiple linear regression was used with 9 features to find the
best features to dive into with more detail. The four features selected from this
analysis were GDP, Stringency Index, Median Age, and Life Expectancy. These features
were analyzed for three countries in each continent to find patterns. It was found
that in the three richest continents GDP and Stringency Index were all positive, while
in the three poorest continents, the coefficients of these features were negative.
This paper assumes basic conceptual knowledge of machine learning and should be readable
by any upper level computer science undergraduate student. Advisors(s): Dr. Mohammed Aledhari maledhar@kennesaw.edu Topic(s): Data/Data Analytics CS 4267Department: Computer Science Presentation|Poster
Judges: Yingying Kang, Tammy Schopf UC-11 Information Recall For Kids With Autism(Undergraduate Capstone)) byAlex J Bechke, Elizabeth Burnside, Haiden Gembinski,Ross Murphy, Ryan Taylor, Henok Demisse Abstract:Description: Our project "Information Recall For Kids With Autism" also known as the
product name given by the client "Safe Kid" is an app to help children with autism
understand basic contact information such as phone numbers, addresses, and names.
This app is being created for our client Spectrum Behavioral Associates who specialize
in helping kids and young adults who have autism, learning development delays, or
other behavioral challenges. Motivation: To teach kids who have autism basic contact
information in case of an emergency. Materials and Methods: The app we are creating
is a local based app made within Unity. The coding language we are using is C# and
we are incorporating user centered design to best fit the target audience. Preliminary
Results: To change a life for someone who has autism or a friend/family member of
someone who has autism. Intellectual or Business Merit: To teach kids basic contact
information through sequences and other methods or recalling information. Also for
us it would be implementing what we have learned through the years as software engineering
students via documentation, user centered design, testing, consulting, and implementing.
Actions That Enhance the Potential of Our Projects Benefit to Society: We feel as
it is very important to learn basic contact information in case of an emergency. Some
kids with autism or other behavioral delays may have a hard time vocalizing the contact
information but have no problem writing it down. This could change the lives of many
kids and family members of children with autism. Advisors(s): Professor - Dr. Parizi Client - Spectrum Behavioral Associates Personnel
- Hannah Taylor Topic(s): Games SWE 4724Department: Software Engineering and Game Design and Development Presentation|Poster
Judges: Yingying Kang, Tammy Schopf UC-12 Comprehensive Security Solution for small E-commerce Business(Undergraduate Capstone)) byPatrick McCollums,Hristo Bakalov,Philinda Morse,Tyler Phillips,Watson Day Abstract:Project Description: Create an e-commerce server and a comprehensive security program
to protect a web server for a simulated small business. This server will include security
tools such as intrusion detection, firewall, and network monitoring. The installation
and maintenance of this solution will be documented as part of the final documentation
package. The server will be reviewed for exploitation from other teams while we attempt
the exploitation of their server(s). Research/Motivation: How to research, install,
configure, and integrate various open-source software packages for information security,
e-commerce, web hosting, and database. Our motivation for this project was to create
and secure an e-commerce website that allows the team to explore, learn, and gain
knowledge to become better real world IT professionals. Materials/Methods Our team
leveraged the use of their own virtual machines and online documentation to test various
software packages on the Ubuntu operating system. We leveraged the NIST cybersecurity
framework to integrate industry standards and best practices to create risk assessment
and information security documents. Preliminary Results: We have created a secure
Internet facing e-commerce solution with supporting documentation. We are currently
awaiting other teams to begin penetration testing and results from of our server.
Intellectual or business merits of our project: Our team gained real world knowledge
and skills during the research and implementation of the server and security project.
Our documentation details the steps taken throughout the implementation of the project
and allows us to hand off the ongoing maintenance to an e-commerce business. Actions
that we'll take to enhance the potential of the project to benefit society: Our documentation
of the project could be published to allow e-commerce businesses to create a low cost,
secure e-commerce store. Advisors(s): Project Sponsor: Dr. Lei Li Professor: Dr. Ying Xie Topic(s): Security IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Piyush Dhawan, Fabio Valbuena, Daniel Omuto UC-16 Understanding the Drivers of Medication Nonadherence in the United States(Undergraduate Capstone)) by Roman A Schwieterman, Austin Poole, Austin Kay, Usman
Mustafa, Issifou Ali Abstract:Medication nonadherence is generally defined as a patient’s inability to take their
medications correctly as prescribed by their doctors. Medication nonadherence adversely
affects patient outcomes and increases healthcare costs. Prior research found that
health system-, condition-, patient- (older age is one factor), therapy- and social/economic-related
factors have been identified to show effect on non-adherence. Our goal is to analyze
the NHIS data to understand the sociodemographic and health causes of medication nonadherence,
as well as answer the following questions about our selected topic: What variables
are the most relevant drivers of nonadherence? Does the direction and strength of
the associations between these variables and medication nonadherence vary over time?
If so, how? Are trends getting better or worse? Do the results of your analysis suggest
that social inequality factors may be linked to medication nonadherence? If so, how?
What are the implications of your analysis for various stakeholders? How does this
vary depending on whether medication nonadherence is intentional or unintentional? Advisors(s): Advisor/Instructor: Ying Xie | yxie2@kennesaw.edu Project Owner/Sponsor:
Dr. Chi Zhang | czhang4@kennesaw.edu Topic(s): Data/Data Analytics IT4893 - IT Capstone (W01)Department: Information Technology Presentation|Poster|More Information
Judges: Piyush Dhawan, Fabio Valbuena, Daniel Omuto UC-19 Comparison of Active and Passive Attention Based Tasks Using EEG with Convolutional
Neural Network(Undergraduate Capstone)) by Jasmine Hemphill, Alyssa Myers, Matt K Warman Abstract:When considering a student's attentiveness while taking online courses, it is known
that they tend to lose focus or get distracted at some point during the lecture. It
is said that as humans we are supposed to learn in active environments. Watching a
lecture from a screen is considered a passive task. Combining that with another factor
like being tired decreases attention even more. Conducting active and passive attention-based
trials will reveal varying results in different states of attentiveness. This project
compares active and passive attention trial results in two states, wide awake and
tired. This has been done in order to answer the questions: Do subjects perform better
(as in maintain concentration and attentiveness) on active tasks while both tired
and awake? And do they perform worse on passive tasks when tired and better when awake?
The data analyzed was collected from electroencephalogram (EEG) waves, and then later
processed through a 3D Convolutional Neural Network (CNN) to produce results. Three
passive attention trials and three active attention trials were performed on seven
subjects, while they were wide awake and again when they were tired. Advisors(s): Dr. Ying Xie Topic(s): Data/Data Analytics IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Piyush Dhawan, Fabio Valbuena, Daniel Omuto UC-20 Analyzing Concentration Levels in Online Education Using Machine Learning(Undergraduate Capstone)) byGray Daugherty,Rachel W Lawson,Kyle Ensley C Ensley,Michael Chann,Tobi Adams Abstract:These past few years have introduced the most important time in history to study new
faucets of online learning. Due to COVID's impact, online learning became a staple
in millions of homes across the country. Guided by the research question, “Can a machine
learning model be created and trained to detect student concentration level based
on eye and facial data?”, we set out to contribute to society’s understanding of online
learning under the guidance of Dr. Ying Xie and Dr. Linh Le. Our process involved
recording ourselves participating in online classes to garnish eye and facial data.
Each student recorded data where they were staring directly at the screen and focused
while also recording data from when the student was distracted and looking elsewhere.
Next, deepfake technology was used to swap our faces with celebrity images to protect
our privacy while also generating a large pool of data. Finally, we created, trained,
and tested different machine learning models to try and find settings best suited
for our needs. Preliminary results indicated our top constructed models hovering around
65-75% accuracy. This was all completed using open source software, and we believe
that if access was granted to even more accurate eye tracking/deepfake technologies,
the accuracy of the model would increase even further. Overall, our results indicate
that it would be possible to develop a model that could indicate whether a student
was concentrated or distracted based on eye and facial data while implementing a process
that could protect the student’s identity, which could be useful data for either the
student to review their own performance in class or for a teacher to see at what point
students become distracted during an online session. Advisors(s): Dr. Ying Xie (Co-sponsor) Dr. Linh Le (Co-sponsor) Topic(s): Data/Data Analytics IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Stephanie Herring, Suneel Mendiratta UC-25 Woodline Interiors Project Planning Application(Undergraduate Capstone)) by Beniamin Costea, Iram Nawaz, Natan Beraki, Daniel Lopez,
Carter Richter, Yeonkuk Woo Abstract:Meeting with the client regularly, we’ve established one of the many solutions to
their requirements. The client, WoodLine Interiors, needs a well-designed system that
can solve product management, client communication, and project management. The main
issue that the company is facing is in the area of managing multiple projects and
tracking project progress. Their issue is commonly addressed by a large number of
companies that create software solutions, but it’s not personalized for their appropriate
needs in the field of cabinetry. As a result our solution created the perfect stages
and states of the project so there is no more confusion to where the project is situation
at. Advisors(s): supervisor : Dr. Reza Parizi project owner : Woodline Interiors (client) Topic(s): Software Engineering SWE 4724Department: Software Engineering and Game Design and Development Presentation|Poster
Judges: Stephanie Herring, Suneel Mendiratta UC-26 Pose Extraction for Real-time Workout Assist_Capstone Group_W01_Spring Semester(Undergraduate Capstone)) byRoyce Camp,Zach Christmas,Jonathon Segars,Amanda Mead,Cameron Page Abstract:Motion Intelligence Research Project We installed and tested many existing pose extraction
technologies in many situations. We provided reports on the different software solutions
and decided on a single solution that performed the best. We will extract key-points
and track the movements across multiple dimensions. We will demonstrate the X and
Y movements of everyone for our chosen software solution in a Business Intelligent
tool (PowerBI). Given the current epidemic, we are not going to be able to compare
against the professional KinaTrax software on campus. Although, Dr. Xie has given
us other software packages to compare our given software solution against. We will
compare the accuracy, speed, and performance of these solutions against ours and demonstrate
the differences in our Business Intelligent tool (powerBI). We will them compile all
our findings in a PowerPoint presentation and Final Project Package. We hope our findings
will assist in Motion intelligence and body tracking research and development. Advisors(s): Dr. Ying Xie & Dr. Linh Le Dr. John Johnson – Exercise Science Topic(s): Artificial Intelligence IT4983Department: Information Technology Presentation|Poster
Judges: Stephanie Herring, Suneel Mendiratta UC-30 Malware Analysis Using Reverse Engineering(Undergraduate Capstone)) byWilliam K Pharr,Kelton Reid,Icyss M Strong,Michael R Lewis, Momodou Mbye Abstract:Cybercrimes are a billion-dollar industry that is rapidly growing by the day. One
of the biggest threats faced by companies is the infection of malware. New forms of
malware are created daily and ever evolving to evade detection methods. Understanding
how malware infects your system and how it eludes detection is crucial to keeping
a company's network and devices safe. During this project we will be using reverse
engineering methods to better understand the functionality of malware, as well as
how it eludes detection. We will be using IDAPro and WiDbg to perform the reverse
engineering. Using this knowledge, we will create a set of security standards to help
companies to protect themselves from these infections. We will also create a document
on how to secure a virtual machine for malware analysis. This will help future students
who also are interested in analyzing malware themselves. Our preliminary results include
understanding some of the most used forms of malware evasion techniques. These techniques
include stalling delays, which is when a piece of malware remains idle to defeat time-based
antivirus scans. Another technique is action required delays, which is when a piece
of malware will only execute once an action or group of actions are performed this
will trigger the malware to execute. Another way that malware is able to evade detection
is fragmentation. In this technique the malware will split into multiple different
fragments, which alone do not raise flags as suspicious, then rejoin and execute. Advisors(s): Dr. Hossain Shahriar Topic(s): Security IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Joe Spalla, Robert Fortunato, Yola Belisario UC-35 Development of an Automated Software Packaging Solution for Linux(Undergraduate Capstone)) byRobert D Ryan,Samantha R Figueroa,Dylan Parker,Blair Hill,Bishwo R Marhatta Abstract:The main problem with using Linux software in the science and Bioinformatics community
is because Linux has a large number of distributions and dependencies. This hinders
researches and science students with the problem of tracking down dependencies for
software which could then further break the existing system dependencies. Our team
looked to solve these problems by creating a BASH script that could quickly mass package
AppImages and contain Linux software with all dependencies. Our team worked through
the last ten weeks and researched all components of AppImage and discovered all means
to more easily package and have a repeatable process for batch software processing.
Our group was able to identify challenges and problems and produced working scripts
to solve our problems. Our results also led us to reach out to the creators of AppImage
and prompt them for future roadmap items and shortcomings of the runtime software
solution. We met our current objectives by producing a BASH script for automated packaging
and proposed ideas for future research in this process. This project will ultimately
contribute to easier consumption of science software for students and researchers. Advisors(s): Capstone Course Instructor: Dr. Ming Yang, myang8@kennesaw.edu Project
Sponsor: Dr. Tsai-Tien Tseng, ttseng@kennesaw.edu Topic(s): Software Engineering IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Joe Spalla, Robert Fortunato, Yola Belisario UC-36 Using Machine Learning Techniques to Predict RT-PCR Results for COVID-19 Patients.(Undergraduate Capstone)) byBradley T Durden,Mathew Shulman,Andy Reynolds,Thomas A Phillips,Demontae Moore,Indya Andrews Abstract:With the COVID-19 pandemic still a threat, healthcare professionals and medical industries
keep searching for better ways to mitigate the spread of COVID-19. While Machine Learning
has been applied in many other domains, there is now a high demand for diagnosis systems
that utilize Machine Learning techniques in the healthcare domain and in particular
combating COVID-19. In this project, we explore the role of Machine Learning models
in combating COVID-19, using WEKA as the main tool for analysis. Advisors(s): Dr. Ming Yang - IT 4983 Capstone Professor Dr. Seyedamin Pouriyeh - Project
Owner Topic(s): Data/Data Analytics IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Joe Spalla, Robert Fortunato, Yola Belisario UC-39 Journalistic Integrity vis Artifical Intelligence(Undergraduate Capstone)) byDylan Dalton,Ray Martin,Duy Nquyen,Yesse Quezada,Brian Dominguez Abstract:We are developing a web app to recognize and rate political bias in online journalism
using artificial intelligence. All human writing inherently contains bias ,however
bias is less harmful if it is transparent to the reader because they can now make
informed decisions about what they read. We've collected articles and and reactions
to them from online sources, and then used Neural Networks trained for natural language
processing to determine bias. The project can predict bias labels on a news articles
with 82% accuracy. Advisors(s): Reza Meimandi Parizi - course instructor Asher Nuckolls - project owner Topic(s): Artificial Intelligence SWE4724Department: Software Engineering and Game Design and Development Presentation|Poster|More Information
Judges: Will Tartak, Bob Cole UR-31 An Empirical Study of Thermal Attacks on Edge Platforms(Undergraduate Research) by Tyler Holmes, Justin Duchatellier Abstract:Cloud-edge systems are vulnerable to thermal attacks as the increased energy consumption
may remain undetected, while occurring alongside normal, CPU-intensive applications.
The purpose of our research is to study thermal effects on modern edge systems. We
also analyze how performance is affected from the increased heat and identify preventative
measures. We speculate that due to the technology being a recent innovation, research
on cloud-edge devices and thermal attacks is scarce. Other research focuses on server
systems rather than edge platforms. In our paper, we use a Raspberry Pi 4 and a CPU-intensive
application to represent thermal attacks on cloud-edge systems. We performed several
experiments with the Raspberry Pi 4 and used stress-ng, a benchmarking tool available
on Linux distributions, to simulate the attacks. The resulting effects displayed drastic
increases in the temperature and power consumption. The key impact of our research
is to highlight the following risks and mitigation plans: the vulnerability of cloud-edge
systems from thermal attacks, the capability for the attacks to go unnoticed, to further
the understanding of edge devices as well as the prevention of these attacks. Advisors(s): Dr. Kun Suo Topic(s): Security Department: Computer Science Presentation|Poster
Judges: Will Tartak, Bob Cole UC-43 Text Marks the Spot Mobile Application(Undergraduate Capstone)) byAustin T Hogan Abstract:A user-to-user location-based messaging mobile application. Ever wanted to leave a
location-specific message to a friend or family member? Leave a grocery list at your
local grocery store for your spouse? Text Marks the Spot provides that functionality
for these use cases & many others. The application will be available for both iOS
& Android mobile device users. Advisors(s): Dr. Reza Parizi - SWE Capstone Course Instructor Topic(s): Software Engineering SWE 4724Department: Software Engineering and Game Design and Development Presentation|Poster|More Information
Judges: Will Tartak, Bob Cole UC-56 Rendeview(Undergraduate Capstone)) by James C Noltimier, Gyasi Igyan, Barrett Rose, Niyi Adekunle,
Ashley Lowe Abstract:Rendeview is a mobile application designed to allow users to find a physical meeting
location equitable for 3+ people, taking into account drive time and traffic conditions. Advisors(s): Dr. Reza Parizi Topic(s): Software Engineering SWE 4724Department: Software Engineering and Game Design and Development Presentation|Poster
Judges: Bruce Skillin, Harrison Wittenbrook UR-41 The Accessibility of the Mobile Gaming Platform for the Visually Impaired(Undergraduate Research) by Christian Thomas Jansen Abstract:The motivation for this project is to research mobile gaming interfaces with the goal
of conceptualizing practices in game design that would create more accessible interfaces
for the visual impairment community. Thus far, the project has focused on practices
that mobile game designers can use to make their games more accessible to the visually
impaired. These includes the use of plain text rather than graphics to be scannable
by screen readers, the inclusion of audio-oriented support and instruction, the use
of contrasting colors to make options more recognizable to those with partial visual
impairments, and the implementation of game mechanics that can be learned and operated
either partially or entirely through non-visual means. Advisors(s): Professor Nicholas Murphy Topic(s): Games CGDD 3103Department: Software Engineering and Game Design and Development Presentation|Poster
Judges: Bruce Skillin, Harrison Wittenbrook UR-66 Image Segmentation with Machine Learning(Undergraduate Research) by Kedar A Johnson Abstract:An experiment-based analysis of the performance of machine learning algorithms in
image segmentation. The experiment is organized to test three experimental groups
representing supervised, unsupervised and reinforcement machine learning. The three
experimental groups are exposed to three datasets of images for training and testing.
They’re performance results are recorded and compared for a statistically significant
difference in mean performance values. These results are assumed to identify a trend
in differences in performance if a statistically significant difference in performance
statistics is discovered between any of the three groups. This experiment will follow
a quasi-experimental design because of the absence of a control group. Advisors(s): Dr. Dan Lo Topic(s): Artificial Intelligence n/aDepartment: Computer Science Presentation|Poster
Judges: Bruce Skillin, Harrison Wittenbrook UC-37 Interactive PDF File Editing for Online Classes(Undergraduate Capstone)) byDavid J Hall,Chris J Stubbs,Justin Masters,Dalton G Parker,Rosendo Lopez Abstract:This system aims to create an interactive environment for teachers to view/grade/edit
student submission in virtual classes. Objectives for this project are to create independent
component or logic model that includes the following functions. This component should
be integrated with a .net core application easily. -Upload pdf files to the system
and save files to the server; -Record audio online and save audio to the system; also,
the audio can be played online; -Upload and play video or video link (YouTube); -Split
file. When uploading a PDF file, the system will allow to split or crop the file (partial
file content) and upload the file; -PDF edit: be able to view the pdf file and leave
comments; Advisors(s): Yang Ming - Capstone Professor Derek Shi - Project Sponsor Topic(s): Software Engineering IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Joe Drury, Amer Uttamchadani UC-57 Cultiva- The Plant Companion(Undergraduate Capstone)) byTravis R Hescox,Kat Branham,Ahsan Jamal,Joseph Henggeler,Erick Reyes,Andy Alcaraz,Momodou Mbye Abstract:The goal of this project is to improve the lives of gardeners or everyday people with
a green thumb by providing them with a planter that will not only hold their plant
of choice, but also give them information about the health and growth of their plant
without them having to interact with it directly. This project will provide the user
with a planter containing sensors that communicate with an application from which
the user can monitor the environment of the plant. This project can serve a wide range
of people from first-time gardeners to seasoned veterans. Users will no longer have
to guess about the health and conditions of their plant, all of this information will
be available to them through a web application. The plant companion will collect data
using sensors that monitor soil moisture levels, light, water reservoir level, water
reservoir pH, and environment temperature. If water, moisture levels, sunlight exposure,
or temperature are outside of the optimal range, the user will be notified both through
the web application. The planter will water the plant through a pump, if there are
sufficient levels in the water reservoir, whenever moisture levels are too low. Upon
logging into their account, users will be greeted with a dashboard where they can
view the data collected by their Plant Companion. Advisors(s): Dr. Ken Hoganson Topic(s): IoT/Cloud/Networking CS 4850Department: Computer Science Presentation|Poster|More Information
Judges: Joe Drury, Amer Uttamchadani UC-62 Machine Learning: Twitter Bots in Disguise(Undergraduate Capstone)) byMatthew Joseph Scheer,Nicolas Vasquez,James C Andersen,Joshua Tiangco,Justin Van,Cody R Walicek,Daniel Rimmel Abstract:This project was designed to help fight against misinformation spread by bots(computers),
the goal assigned to us was to find and inform Twitter users of bots that follow and
are being followed by the user. Advisors(s): Dr. Reza Parizi Topic(s): Artificial Intelligence SWE 4724Department: Software Engineering and Game Design and Development Presentation|Poster|More Information
Judges: Joe Drury, Amer Uttamchadani UC-69 Team 10B BChain: secure peer to peer file sharing(Undergraduate Capstone)) byJonathan D Lashgari,Carlos A Diaz,Jeffery Erhunse,Caleb T Goff,Giang T Nguyen Abstract:BChain is a new P2P file sharing system that is fully private, anonymous, globally
self-verifying, and utilizes an automatic peer-maintained network of trust in data,
accomplished through new methods of routing content over the whole network, encrypted,
rather than per torrent download. Verification is done by adding file metadata to
a blockchain giving the network consistent knowledge of each file it can transfer,
and how to verify file received against the network. This enables a policy of zero
trust against peers. This system is implemented by an app that interfaces with the
network using the protocol, using it for upload, download and file discovery. The
interface is built using web technologies, which allows for flexible use across native
platforms. Advisors(s): Prof. Ken Hoganson Topic(s): IoT/Cloud/Networking CS 4850Department: Computer Science Presentation|Poster|More Information
Judges: Ryan Taylor, Leonard Greski UR-46 BreastNet;(Undergraduate Research) byCora L Meador,Ryan Deem Abstract:In the United states, 13% of women are diagnosed with breast cancer in their lifetime,
and it is the second leading cause of death by cancer in women. Early detection and
screening can result in an increase of life expectancy by 10 years on average. Unfortunately,
breast cancer can be challenging to detect, since it can appear anywhere in the breast.
Cancer that is detected in its early stages can give patients more options and save
thousands of dollars in medical costs. Some of the most recent developments in computer
science and machine learning are in the biomedical field, especially individualized
healthcare. There is also an increase in the demand for telehealth options, reducing
healthcare costs. With the help of computational technology, medical practitioners
will be able to process data more quickly, which will allow more patients to have
access to reliable treatment. Besides, systematic processes for interpreting various
data types (such as clinical features, genetic information, and medical images) can
identify trends that a human eye would not detect. This project aims to design and
implement an artificial intelligence-based model called BreastNet to classify breast
cancer into high and low-risk categories based on a combination of MRI images and
clinical data. BreastNet uses a convolutional neural network (CNN), a type of machine
learning methodology that imitates how the human brain learns information. Neurons
fire in a connected pathway, reinforcing the relationship between a stimulus and the
correct outcome. In this case, the CNN identifies characteristic features within the
MRI that correspond to different life expectancy outcomes, which are notated in the
clinical data. The clinical data serves as a loss function, which allows the network
to identify how well the current model performs on images. We will evaluate the model
by dividing the dataset into three partitions: training, validation, and testing,
and then uses the evaluation metrics of Accuracy, Loss, F1 Score, Precision, Recall,
Specificity, and Sensitivity. Advisors(s): Dr. Mohammed Aledhari Topic(s): Artificial Intelligence CS 4267Department: Computer Science Presentation|Poster
Judges: Ryan Taylor, Leonard Greski UR-48 Using Semantic Segmentation in a Convoluted Neural Network for Vocal Localization
in Music(Undergraduate Research) byTrevor E Stanca,Noah Trinite Abstract:I. PROJECT OVERVIEW A. Research Question In this project, the question was asked:
”Is there an easier way to extract vocals from music?” Many other works are able to
extract vocals with Deep Neural Networks using Multitask Learning, which are large
and take a long time to train. To rival this, we wish to present a method to identify
vocals with a Convolutional U-Network (U-Net) for Semantic Segmentation of audio files.
B. Project Description This project differs from other works by identifying vocal
locations by converting audio files in Short Time Fourier Transforms(STFT), and treating
them as images in the UNet. By treating these as images, the U-Net is able to identify
the location of ”vocal features” the same way a U-Net would identify desired features
within an image. The object detection is what sets this project apart from similar
works. Many of these other works treat each song as an audio signal with real and
imaginary components which means these algorithms treat the issue as a signal processing
problem. However, by looking at the STFT of the song as a graph, we are instead able
to approach this as an image processing problem instead, which offers more tools within
the realm of Deep Learning–such as Semantic Segmentation. II. EXPERIMENTATION A. Materials
and Methods All Materials used were a form of software. Firstly, the UNetwork was
created and ran in python on the CCSE Cluster for High Efficiency. A U-Network is
a Convoluted Neural Network that has the ability to output images by Convoluting the
original image to allow only the prominent features to be shown and Deconvoluting
the Output to display these features in the original image resolution to be used for
further processing. This gives the U-net it’s ”u” shape when drawn out. Secondly,
the data created for the project were music files converted into Short Time Fourier
Transforms(STFT) and processed as image files, where the input into the U-Network
was an entire song’s STFT and the labeled data was the vocal audio file STFT for that
same song. A Short-Time Fourier Transform can be considered the heatmap of the amplitudes
of the song across frequency and time. B. Results The initial Results from the U-Network
show a high level of accuracy for vocal location predictions. As the output from a
U-Network is an image, these images are the initial song’s STFT with a mask applied
to show the location of Vocal Waves. These trials have an accuracy greater than 80%
which is a very good result this early in the processing. The vocals have been identified
and located in this study, however the next step is to pull the vocals out and convert
them back into a song wave. III. MARKETABILITY For the last 20 or so years, large
record labels have been attempting to ”Remaster” old music, which is the process of
digitizing old analog tracks of songs, mixing them on a new sound board, and releasing
the remastered work at a marked up price. As recording methods, pre-computers, relied
on tape, often times tracks were record over each other to save space on the real.
When the song has this issue, a computer program has to pull out all of the pieces
of the song so that the engineer can remaster it. This project shows the initial steps
to a simpler audio extraction, where handling this issue as an image processing problem
instead of a signal processing problem, we are able to create a more efficient Neural
Network. Advisors(s): Dr. Aledhari Topic(s): Artificial Intelligence CS 4267Department: Software Engineering and Game Design and Development Presentation|Poster
Judges: Ryan Taylor, Leonard Greski UC-59 Analyzing Concentration Levels in Online Learning with Facial Values(Undergraduate Capstone)) byElliott J Witherell,Jakeira Askew,Jonathan R Dicks,Steven C McGuire,Jacob A Walton Abstract:Can deep learning models accurately predict whether an individual is focused or distracted
on a task in order to improve learning efficiency? In the context of online learning
with the use of a webcam, this project is aimed at detecting concentration levels
of students to potentially assist with improving learning efficiency. Machine learning
technologies have been utilized to evaluate students’ facial expression and eye movements
to identify whether a student is focused or distracted. The machine learning branch
that is employed is a supervised learning model. This supervised learning model makes
predictions based on given input features. A total of 6 different models were employed.
4 of those models employed collected eye data. The other two models employed the use
of facial and eye data to predict concentration. Ultimately, the eye model accuracy
hovered between 50% and 56% accuracy in prediction, with a significant amount of loss.
The eye models with attention provided the best accuracy and loss rates out of the
four eye models. Secondly, the facial and eye models also hovered right around 50%
accuracy with significant loss of around 3.8 and 3.7. The reported results suggest
that the data was inaccurate or insufficient in some models to accurately predict
concentration levels in an individual. Given a larger collection and more consistent
data, the reported results would provide to be more accurate at predicting concentration. Advisors(s): Dr. Linh Le (Sponsor/Project Owner) Dr. Ying Xie (Sponsor/Project Owner) Topic(s): Artificial Intelligence IT 4983Department: Information Technology Presentation|Poster|More Information
Judges: Bill Forsyth, Abdul Wahab UR-60 Video-to-Video Synthesis With Semantically Segmented Video(Undergraduate Research) byAydan Mufti,Jordan S Hasty Abstract:Our project involves studying the usage of generative adversarial networks (GANs)
to translate semantically segmented video to photo-realistic video in a process known
as video-to-video synthesis. The model is able to learn a mapping from semantically
segmented masks to real-life images which depict the corresponding semantic labels.
To achieve this, we employ a conditional GAN-based learning method that produces output
conditionally based on the source video to be translated. Our model is capable of
synthesizing a translated video, given semantically labeled video, that resembles
real video by accurately replicating low-frequency details from the source. Advisors(s): Dr. Mohammed Aledhari Topic(s): Artificial Intelligence CS 4732Department: Computer Science Presentation|Poster
Judges: Bill Forsyth, Abdul Wahab UR-63 Low Cost High Impact Fall Detection At The Edge(Undergraduate Research) by Dylan SIRNA Abstract:ML models have become more accurate, powerful and portable in recent years, the purpose
of this project is to explore how these advances can be applied towards fall detection
for less cost than before possible. This project explores the application of micro
controllers which have become cheaper and stronger along with emerging machine learning
models that can be trained on a traditional computer with greater resources and then
port the model to be interpreted on a micro-controller such as a raspberry pi. These
two factors lead to the reason to revisit the problem of fall detection, a problem
that plagues the elderly can likely be solved cheaper and more accurately than ever
before, and that is the challenge that this paper aims to explore. Advisors(s): Professor Mohammed Aledhari Topic(s): Artificial Intelligence CS 4267Department: Computer Science Presentation|Poster
Judges: Bill Forsyth, Abdul Wahab UR-65 CNN CIFAR Image Identification(Undergraduate Research) by Matteo L Staciarine Abstract:Reducing the learning rate of a CNN can positively affect the validation accuracy
of a machine learning model. Dropping out nodes from different layers can further
delay overfitting from happening. Validation loss decreases over more epochs, but
it must be cut when it reaches its minimum value. Advisors(s): Dr. Dan Lo Topic(s): Artificial Intelligence Department: Computer Science Presentation|Poster
