A Snapshot of Summer Research Projects

With the Fall 2017 semester now underway, we take a look back at faculty and students who have continued their education throughout the summer at Harbor Walk. Research with our faculty is just one of many opportunities students have access to within the Computer Science department at the College of Charleston. Below are a few of the students who took advantage of this wonderful opportunity to do research over the summer break.

Three students worked in the Cyber Security Lab under the guidance of Dr. Xenia Mountrouidou. Their topics revolved around Internet and Android security, as well as cybersecurity education.

Blaine Billings, an honors student that just completed his freshman year as a double major in Computer Science and Mathematics, worked on Internet of Things (IoT), taxonomies, and models of operation. The goal of his research is to achieve a deep understanding of the ecosystem of IoT devices through a taxonomy that groups them based on their specifications and behavior. Based on this general description of devices, Blaine is currently working on probabilistic mathematical modeling of the correct behavior of web-cameras, one of the most vulnerable devices that has been manipulated in large botnets. The goal is to develop a situational awareness for Internet connected devices and detect anomalies in operation when these devices are used in botnets that threaten name servers and other fundamental network operations. Blaine’s work will be presented at the SIGCSE ACM undergraduate research competition on February 2018. In addition, Blaine worked on cybersecurity education materials. He completed the development of a lab where students can practice privilege escalation in an isolated cloud environment, the Global Environment for Network Innovations (GENI).

Daniel Baczmaga, a rising junior Computer Science major and an honors student, worked on security assessment of digital assistants. Specifically, Daniel performed a network vulnerability assessment for the Amazon Echo digital assistant and will expand on Google home. He has gained insight on the network handshake that the assistants perform with the cloud and mobile phone, with goal to complete a man in the middle attack to intercept or interfere with communications.

Louis Mejia is one of our graduate students in Computer and Information Sciences that just completed his first semester at the College of Charleston. His work focuses on Android API vulnerability assessment using fuzzing techniques. Louis is working on finding fundamental flaws in the API that can affect any mobile application that uses specific libraries. He is focusing on testing the Bluetooth and storage libraries. These libraries touch a large majority of applications that connect via Bluetooth to medical or home devices and affect the storage of personal identification data. He has also set up an experimental IoT network for security assessments.

 

For the third summer, Dr. Paul Anderson led a REU host site which was sponsored by the NSF: http://omics.cofc.edu/.  Several students performed research over the summer on a variety of data science related projects. The REU included everything from using machine learning to determine biomarkers for lung cancer relapse, to using next-generation sequencing methods to study the genomics of sea turtles, to advancing big data processing architectures such as Apache Spark for big data genomics, to developing novel web applications for breast cancer research, to developing software for STEM education. Final presentation day on August 3rd, 2017 featured the following talks:

 

  • Nia Kyritsis and Micah Mills: RNAseq analysis of temperature controlled brain and gonad gene expression provides insight into potential selection on transcripts linked to life history adaptation and environmental change in Loggerhead sea turtles (Caretta caretta)
  • Abigail Moore and Brandon Zheng: RNA-Seq analysis to detect biomarkers in non-small cell lung cancer patients
  • Caroline Oliver: Complementary data analysis of phosphoproteomic and proteomic data for ovarian cancer patients
  • Katie Duchinski: SLKBase: A Comprehensive Online Database for SUM Breast Cancer Cell Lines
  • Elise Burton: Large Data Processing Pipelines for Genomic Data
  • Larry Davis: Big Data Genomics: Apache Spark-based distributed bioinformatics pipeline
  • Jesse Mayeaux: Exploring the NIST Bottlenose Dolphin Genome Assembly as a Resource for Teaching Comparative Evolutionary Genomics
  • Brooke Gantt: Human-Dolphin Genome App for interactive learning in the STEM curriculum

 

Four students worked with Dr. William Bares in the Virtual Production Lab. Romain Simon was an internship student from La Rochelle, France (April – July). He explored machine learning techniques in Python including clustering and natural language processing to learn models of cinematography.  These Python prototypes will also be used to develop course materials that use data science techniques in applications of Computing in the Arts. Rex Ferrer, computer science  major, used the Unity Game Engine to animate 3D characters and visualize their motion paths. Elizabeth Obisesan, an exchange student from Nigeria, developed a JavaScript prototype to be used in automating the creation of questions about movie scenes for use in conducting crowd-sourced surveys. Yasmin Coker, a summer high school intern from SC Governor’s School for Science and Math, used Blender 3D to create a virtual 3D reconstruction of the character blockings for a dialog scene between Gandalf and Elrond from Lord of the Rings.  She also conducted a literature search to produce an annotated bibliography of research works in crowdsourcing related to movies.

 

Four students worked in the UAV Robotics lab with Dr. Sebastian van Delden: Eduardo Abreu (Computer Science major at CofC), Erin Puckette (Aerospace Engineering major at UVA), Joseph Puryear (Aerospace Engineering major at Princeton), and Iain MacIver (Mechanical Engineering major at Clemson). These students worked on two disparate projects. The first involves having a drone fly an autonomous flight pathway in a vertical grid pattern so that a image mosaic or 3D model of a vertically oriented surface could be reconstructed: https://youtu.be/G8x1dCRCrk8. The system could be used by geologists studying cliff faces or historians studying architectural structures. The second involves programming Epson Moverio Smart Glasses to control a drone using head movements while the drone’s video stream is shown on the heads-up display. This system would provide a hands-free and immersive approach to flying a drone.

 

 

Paige Peck (CITA major at CofC) and Larr Brock (Cybersecurity Graduate Certificate at CofC/the Citadel) worked on ensuring correctness in healthcare billing under the guidance of Dr. Aspen Olmsted. Healthcare medical billing has been progressing into the digital era for several years, but it has been a slow and expensive process that has left many parts of the industry behind. One of the many things that have been overlooked in the progression is security, especially now that medical records are worth far more than credit card numbers on the black market. Another issue the healthcare industry has been dealing with is the lack of systems being incorporated. Currently, there are companies that are using printed out spreadsheets to find rules for coverage of a procedure based on any insurance company’s policies. Using business rules engines and rule validations, we make it easier for a doctor or office to type in lab results and see whether a procedure will be covered by a patient’s insurance company. The students produced a paper that concluded their research.

 

Phillip Byrd and Daniel Lee worked under the supervision of Dr. Ayman Hajja on developing a web application to extract object-driven action rules from information systems. Action rules are actionable tasks that describe possible transitions of instances from one state to another with respect to a distinguished attribute, called the decision attribute. For example, action rules can help physicians by providing them with suggestions for actions (e.g. change in course of treatment, or dose changes) that will yield desired changes on the patient’s overall condition. Object-driven action rules extraction is a special case of extracting action rules by adapting the techniques used in the classical action rules approach to optimize and account for the limitations and constraints of information systems that exhibit an object-driven nature, which are systems that contain multiple objects, where for each object there exists multiple observations; a typical example of an object-driven system would be a system containing information about multiple patients, where each unique patient is considered an object, and where for each patient there are multiple recorded visits. In addition to building the web application to extract object-driven action rules, Phillip and Daniel started exploring the potential of expanding on the literature of object-driven action rules by examining the possibilities and benefits of creating sequences of action rules through chaining multiple single action rules extracted from independent objects, for the goal of generating and providing decision makers with more thorough multi-step actionable patterns.