2014 Research Experiences for Undergraduates
PROGRAM OVERVIEW
The Team for Research in Ubiquitous Secure Technology sponsored 15 undergraduate students to participate in the summer 2014 TRUST REU program. Below are descriptions of the 2014 TRUST-REU research projects and links to each student's or team's research report and poster presentation.
RESEARCH PROJECTS
Ibrahim Altaweel
Diablo Valley College
Web Privacy Census: HTML5 Storage Takes the Spotlight As Flash Returns
Building on previous studies, here we report on the state of internet tracking on the most popular web sites. We found a 38% increase in HTML5 local storage usage on 55 of the top 100 sites, as compared to 34 sites in 2012. We saw an unexpected increase in Flash cookies in 2014. Fortyone of the top 100 Global sites had flash cookies, with 25% originating from Chinese websites. Since the 2012 report, we have also noted a growth in alternative methods, besides HTTP cookies, that trackers have utilized to gather information from unsuspecting users on the internet. Here we compare data with the 2012 Web Privacy Census and discuss the patterns and trends we see surrounding the current state of web privacy.
Jaime Cabrera
California State University, Fullerton
Web Privacy Census: HTML5 Storage Takes the Spotlight As Flash Returns
Building on previous studies, here we report on the state of internet tracking on the most popular web sites. We found a 38% increase in HTML5 local storage usage on 55 of the top 100 sites, as compared to 34 sites in 2012. We saw an unexpected increase in Flash cookies in 2014. Fortyone of the top 100 Global sites had flash cookies, with 25% originating from Chinese websites. Since the 2012 report, we have also noted a growth in alternative methods, besides HTTP cookies, that trackers have utilized to gather information from unsuspecting users on the internet. Here we compare data with the 2012 Web Privacy Census and discuss the patterns and trends we see surrounding the current state of web privacy.
Jared Compiano
University of North Carolina at Chapel Hill
Neural Networks for Improving Wearable Device Security
Wearable devices with first person cameras, such as Google Glass, are becoming ubiquitous. Additionally, the importance of security will increase as these devices are likely to capture private or sensitive photos. In this paper we introduce a Convolutional Neural Network (CNN) which is capable of detecting various security risks with relatively high accuracy. Our approach is also capable of classifying images within approximately half a second-fast enough for a rapid, mobile environment. We also investigate numerous routes for improving accuracy with our CNN classifier including object segmentation and artificial data generation. Finally we verify our approach using a manually collected dataset which was tested using two different Android devices.
Rachel Davis
Drake University
Improving Protein Structure Prediction Utilizing A Reformed Filtering Process
Knowledge of protein structure grants insight on a protein's function and capabilities within a system. Currently available automated prediction of structures is limited and inconsistent. The Critical Assessment of protein Structure Prediction (CASP) is a worldwide competition to automate the process, and supplementary WeFold allows competing groups to connect with ideas on how to predict proteins. By examining a popular filtering process on the WeFold pipeline, a clearer look into one aspect on why prediction is untrustworthy is presented. Comparing the number of accurate models based on a global distance test to a given template before and after the filtering procedure gives a sense of how well the program is filtering. By adjusting the parameters presented in the filtering code, more accurate results can be yielded.
Zackory Erickson
University of Wisconsin
Neural Networks for Improving Wearable Device Security
Wearable devices with first person cameras, such as Google Glass, are becoming ubiquitous. Additionally, the importance of security will increase as these devices are likely to capture private or sensitive photos. In this paper we introduce a Convolutional Neural Network (CNN) which is capable of detecting various security risks with relatively high accuracy. Our approach is also capable of classifying images within approximately half a second-fast enough for a rapid, mobile environment. We also investigate numerous routes for improving accuracy with our CNN classifier including object segmentation and artificial data generation. Finally we verify our approach using a manually collected dataset which was tested using two different Android devices.
Rachel Harred
University of North Carolina, Greensboro
Analysis of NERSC Job Queue Prediction Tools for Efficient Use of Cray Systems
Supercomputers are used to run extremely large calculations for scientific purposes that would take a normal computer months or even years to complete. The NERSC facility, part of the larger Lawrence Berkeley National Laboratory, is one of the largest centers for supercomputing. One challenge with supercomputing is handling a large volume of jobs of different sizes and run times. The queue system currently in place has a program that attempts to predict how long each job will wait in the queue before starting. This program, called "showstart", is not very accurate and can frustrate users. This research project attempted to discern how showstart made its predictions and to see if a better way was possible.
Kathryn Ho
Mount Holyoke College
Web Privacy Census: HTML5 Storage Takes the Spotlight As Flash Returns
Building on previous studies, here we report on the state of internet tracking on the most popular web sites. We found a 38% increase in HTML5 local storage usage on 55 of the top 100 sites, as compared to 34 sites in 2012. We saw an unexpected increase in Flash cookies in 2014. Fortyone of the top 100 Global sites had flash cookies, with 25% originating from Chinese websites. Since the 2012 report, we have also noted a growth in alternative methods, besides HTTP cookies, that trackers have utilized to gather information from unsuspecting users on the internet. Here we compare data with the 2012 Web Privacy Census and discuss the patterns and trends we see surrounding the current state of web privacy.
Axenya Kachen
University of California, Berkeley
Title of Paper
Abstract
Yuen Wan Lee
Napa Valley College
Custom accounting for the Simple Linux Utility for Resource Managementr
An efficient way to manage the workload on High Performance Computing Clusters is to use Resource Managers. One such of the resource manager and job scheduling system for Linux clusters is the Simple Linux Utility for Resource Management (SLURM). At National Energy Research Scientific Computing Center (NERSC), users regularly submit and run hundreds of jobs. Thus, it is necessary to have an accounting tool to track their job history. This article describes a new tool to query and display the accounting data in SLURM.
Hen Choi Ortiz
University of California, San Diego
Web Privacy Census: HTML5 Storage Takes the Spotlight As Flash Returns
Building on previous studies, here we report on the state of internet tracking on the most popular web sites. We found a 38% increase in HTML5 local storage usage on 55 of the top 100 sites, as compared to 34 sites in 2012. We saw an unexpected increase in Flash cookies in 2014. Fortyone of the top 100 Global sites had flash cookies, with 25% originating from Chinese websites. Since the 2012 report, we have also noted a growth in alternative methods, besides HTTP cookies, that trackers have utilized to gather information from unsuspecting users on the internet. Here we compare data with the 2012 Web Privacy Census and discuss the patterns and trends we see surrounding the current state of web privacy.
Katherine Sittig-Boyd
Simmons College
SciDB for Metagenome Analysis
Utilizing array databases to store large quantities of metadata and data is growing increasingly more common among scientific users as an alternative to traditional data storage methods, such as file systems or relational databases. Implementing SciDB, an open-source array-based database management system (DBMS), may be more optimal for large-scale data analysis. However, the comparative querying performance of each method requires consideration. In this paper, we investigate the query speed of SciDB in comparison to the SQLite _le system utilized by the researchers at the Joint Genome Institute. Determining comparative performance may lead to improvements in how scientific data is stored and queried for more efficient research practices.
Yegeta Zeleke
University of California, Santa Cruz
Title of Paper
Abstract