Department of Computer Science Links/Abstracts

Grayson Buchholz, Joe Pravder

Project Link: https://tulane.box.com/s/1vdav714dqapnz7msk7w9moxvyeuofhf

Through leveraging a vulnerability of pro_ftpd 1.3.5, we were able to install a backdoor on a Metasploitable3 instance within AWS. This attack involves steganographic techniques to evade IDS and SIEM detection. We also used advanced scripts to add privileged users to the /etc/passwd file and delete log traces. The defense solution combines code obfuscation, an IDS detection rule, and moving target defense of the /etc/passwd file.

Mentor: Dr. Zizhan Zheng

Mary Pwint

Project Link: https://tulane.box.com/s/4rfm0fkeag357g0gp86zmohpoc8sfet5

COVID-19 is caused by the pathogen known as Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2). According to WHO’s COVID-19 dashboard, as of April 7, 2021, there have been over 2 million deaths reported worldwide. The mortality of COVID-19 has been largely attributed to cytokine storm syndrome, which is an unregulated hyperinflammatory response due to the systemic spread of a localized inflammatory response. In this project, an agent-based model (ABM) is developed to understand the pathophysiological mechanism of cytokine storm phenomenon and lung damage caused by SARS-CoV-2 exposure.

Mentor: Dr. Michael Mislove

Kyle Strougo, David Ziman, Rebekah Doochin

Project Link: https://tulane.box.com/s/r2ujekhl1lcmjn57b9bny6nlymsqznoy

For our capstone project we decided to see if we could help improve contact tracing tracing around campus in order to mitigate the spread of COVID in and around the Tulane community. Our approach to contact tracing is designed in such a way that would encourage students to work with Tulane’s professional contact tracers rather than feel as though they should hide any events they had attended. We are working with the Assistant Vice President of campus health as well as with Tulane’s IT Architect to create a webpage that could be both easy for students to use as well easily incorporated into the current contact tracing protocol.

Mentor: Dr. Zizhan Zheng

Mark Lisi, Molly Lyons, Dominique Perriseau, Dan Ellsworth

Project Link: https://louisianacoastalrestoration.github.io/index.html

The land of southern Louisiana is eroding at a rate of a football field an hour. Thanks to increasingly frequent hurricanes and storm surges, this rate shows no signs of slowing down. Our project seeks to mitigate this damage as much as possible via the algorithmic placement of cypress trees in Louisianan wetlands. By prioritizing the placement of trees in water rather than on land, our algorithm most effectively protects exposed faces of land and consistently outperforms the human intuition in metrics of land mass retention, as well as adhering to geographic restrictions such as salinity and water depth.

Mentor: Dr. Aaron Maus

Jacob Geisberg

Project Link: https://tulane.box.com/s/ijazeg7yxzwh1q4y8qyfbm08zzno1p33

Since COVID-19 burst onto the international scene in December 2019, researchers from all corners of the globe have been pouring time and resources into studying countless aspects of the disease from its genetic code to its psychological consequences. The unplanned and unprecedented rush to generate and process all that data has led to significant data quality challenges as well and missing resources. My website plugs one of those holes by providing a free, public, and simple way to analyze mutations in COVID 19 genomes through its custom analyses and drag and drop visualizations.

Mentor: Dr. Aaron Maus

Daniel Licht, Marc Ojalvo

Project Link: https://tulane.box.com/s/6rqcfsiscwiy1j2tomjfo4or6x237ydg

Our project aims to determine what role cryptocurrencies will play in the post-pandemic environment and whether the “sector-specific” coin model of using particular coins for different economic purposes is sustainable for the cryptocurrency market. To do this, we have used economic data, COVID-19 data, and cryptocurrency prices from January 2020 to January 2021 to investigate the stability of the cryptocurrency market in the face of recession and global economic crisis. We have utilized various machine learning models, primarily random forest classifiers, and several regression models to gain insight into the viability of cryptocurrencies and the sector-specific coin model for future crises.

Mentor: Dr. Ramgopal Mettu

Lauren Sussman, Leah Kuperman, Jillian Baggett

Project Link: https://tulane.box.com/s/vgd56fgyy29pb9w46wc0lpu5q2r9514j

Our team created a website that helps recommenders spot gender bias in their letters before submission. Past research has shown that when people write about women, they often use more “doubt-raising” language and focus on personality traits rather than tangible accomplishments. We used machine learning and natural language processing to continue this research, and created a website that can identify these areas of bias. Our goal with this project is to bring awareness to gender bias in writing, since it has proven to have harmful effects on the admission of woman applicants to certain opportunities.

Mentor: Dr. Aron Culotta

Matt Catalano, Gabe Harris, Sam Minix, Leo Simanonok

Project Link: https://tulane.box.com/s/3r6d3szjcirmecs16nytip9umkquvpxg

Our goal for this project was to sort comments from articles to improve online discourse. The increased usage of social media to discuss news and politics has also increased distracting and derailing comments from people reacting to headlines. Our project is based on the online forum, reddit.com, which is a popular discussion platform. We use several language processing techniques, including word vectors and named entity recognition. Our final product is a web page that sorts the comments from a reddit link. Users can also input a potential comment to see if it would be good for increasing discussion.

Mentor: Dr. Ramgopal Mettu

Alden Pratt, Daniel Yang, Sam Childs

Project Link: https://tulane.box.com/s/3pwcvsspv6bcco9czhwq25fns7ezvyg3

Investment Groups and Hedge Funds use advanced techniques and proprietary data sets to their advantage when choosing stocks to invest in. Their goal is to drive higher returns on investment without assuming any additional financial risk. Our project investigates whether collecting data from twitter could serve as a publicly available alternative to match the performance of these funds. We collected tweets about publicly traded stocks and market indices and rated them using a sentiment analysis algorithm. In doing so, we created a series of interactive visualizations to track the predictive power of various sentiment indices.

Mentor: Dr. Jihun Hamm

Max Brandell, Madison Davis, Sarper Tutuncuoglu

Project Link: https://tulane.box.com/s/incc7fdkjmkvtmkvevimx2mzinunbe3l

An EEG is a device that reads voltages from the brain through the scalp and compares them to a reference, creating a basic map of activity in the brain of the subject. As the EEG measures what is happening inside the brain and basically outputs raw numbers, it is an excellent candidate for a brain-computer interface. We decided to use this interface in an attempt at improving upon recommendation algorithms by adding an extra type of input representing the emotional state of the user. This will be accomplished by using existing emotionally labeled EEG data from the DEAP dataset to build a machine learning model that can classify a new user’s emotional state from an EEG being performed on them.

Mentor: Dr. Brian Summa

Grace Williams

Project Link: https://tulane.box.com/s/zobnd765mojaypi1ryqzbse55u94i75m

As computer science escalates in educational relevance, accessibility to younger audiences becomes paramount. To increase computational literacy, children need an engaging way to learn these key problem-solving techniques. This project is an interactive graph-traversal visualization that approaches the topic as a puzzle game, aiming to present graph theory as intuitive and accessible so that these principles can be built upon with more sophisticated topics later. The final encoding will have all the characteristics of a puzzle game, such as levels, gamemodes, and scores, while maintaining an emphasis on education.

Mentor: Dr. Carola Wenk

Tullia Glaeser

Project Link: https://tulane.box.com/s/rptqaysda9hmixzjxxe9jv4ymoo2ms1d

As data analysts have a lot of data to sift through, it is important to organize this data well according to the client’s priorities; while indicator scoring itself is not a new concept, it has always been fairly basic. This project provides enhancements to indicator scoring — increasing customizability and transparency — by providing the user with the possibility of scoring different indicator characteristics, displaying these changes to the user through a graph, and allowing the user to push the changes when content with them. A transparent default scoring algorithm is also provided based on the user’s answers to three questions.

Mentors: Dr. Michael Mislove and ThreatQuotient

Riley Juenemann

Project Link: https://tulane.box.com/s/i0qoklmbggj2ltup5yr8xjptp1tgmw6z

Scientists generate hundreds of time series particle trajectories as an output of each fluorescence microscopy experiment. Machine learning provides one opportunity to efficiently process and analyze this large volume of data. Using statistical methods and a support vector machine, we have developed and deployed an accessible, automated tool targeted to the single-particle tracking community for categorizing free diffusing, anchored diffusing, active transport, and subdiffusing particle movement. This involved the generation of simulated data across a wide-ranging, biophysically relevant parameter space. We also compared the performance and relative strengths of supervised and deep learning methods for this problem.

Mentor: Dr. Jihun Hamm

Lindsay Hardy, Samantha Rothman

Project Link: https://tulane.box.com/s/h7a9hashu7fz8sbrad6yiu5eirn2d3o1

Inspired by the events of 2020, we designed a capstone project that will help spread information about protests happening in America. To do so, we used Twitter data to train a classifier that will determine whether a Tweet is about a protest.  The classifier’s results are displayed on a map of the United States that is updated in real time with summaries of the Tweets classified as protests in the past 24 hours. We also include a comparison between our protest data and the data from CountLove and the Crowd Counting Consortium to examine the difference between using social media and the news, respectively, for information about protests. We hope to see people post more on social media about protests throughout the country and that our use of social media to compile information about protests will make it easier for people to understand the protests around them and potentially inspire more.

Mentor: Dr. Aron Culotta

Aline Tran

Project Link: https://tulane.box.com/s/jhjgw14ezebquhgr50a27330jms2zsk3

Double blind studies (where both the researchers and participants are blinded to the allocated treatment) are considered the gold standard in intervention based studies, meaning testing of different treatments for ailments, such as experimental drugs. Randomization produces comparable groups and eliminates the source of bias in treatment assignments, including accidental bias and selection bias. Large studies, such as the Pfizer vaccine study, utilise web based apps for randomization for the convenience of researchers. However, the web apps only use simple randomization or block randomization techniques, which do not take covariates into account. This is inconsequential for large studies due to the sheer number of participants mitigating the effects of covariates, but not ideal for smaller studies. Therefore, there is a need for a web based randomization app that takes covariates into account as well as use different randomization techniques for the convenience of the researchers as well to increase the reliability of the experimental results.

Mentor: Dr. Michael Mislove

Joshua Kellner

Project Link: https://tulane.box.com/s/ylv6bimfsx7zgy3g1jdbh6s0c8x7i04y

This project is a suite of audio plugins that can dynamically react to music being played in real time. It attempts to solve the problem of monotonous loops in electronic music by bringing the reactivity of improvisational music to that medium. The program changes parameters of audio effects based on musical qualities specified by the user. This program is at its full potential when it is using improvised musical input to control the effects of prerecorded/preprogrammed music. This will create a sense of uniformity and dynamic quality that would require much more work to implement in typical electronic music situations and could occur in real time.

Mentors: Dr. Brian Summa and Dr. Rick Snow

Lana Biren

Project Link: https://tulane.box.com/s/ebr3yppgob07c0umfzo58s5vxfy9z89q

For my Capstone project, I aim to help answer the question if algorithms can help mitigate the problem of gerrymandering or create more transparency in the redistricting process as well as test each one’s criteria. My approach is to write an analysis of the outcomes of running different redistricting algorithms on a map of Louisiana using 2010 public census data. I focus on analyzing congressional districts and state senatorial districts in Louisiana. Gerrymandering is a huge problem that affects our democratic process and creates unfair elections. My analysis can raise awareness about how computer algorithms could be a possible solution to gerrymandering.

Mentor: Dr. Ramgopal Mettu

Josh Ballagh, Isabella Casas

Project Link: https://tulane.box.com/s/8a0e3zdl4o8wfwh7wmgzh9ndkcw3255t

Root for Bots is a fully-scripted implementation of the popular asymmetrical strategy board game Root: A Game of Woodland Might and Right by Leder Games that is built to allow AI players. This implementation is for Tabletop Simulator, a game with a very active developer community. We believe that our implementation is an interesting task environment for AI development and will see great use within the TTS community.

Mentor: Dr. Nicholas Mattei

Daniel Margulies, Dave Streit, Jared Blavatt

Project Link: https://tulane.box.com/s/m8kpo00gebpzzgkvqw4f7qlypk49f4d6

We have made the first mobile app that uses computational geometry and machine learning to predict the grade a sports card will receive. Sports cards are more valuable when they are certified mint condition by a professional grading company, but getting a card graded takes time and money, hence the need to have a tool that can predict the grade a card will receive. Our app is very simple, users take a picture of the card, and the image is sent to our machine learning algorithm, which outputs a grade back to the app for the user to see.

Mentor: Dr. Carola Wenk

Ben Nguyen, Garrett van Beek

Project Link: https://tulane.box.com/s/kzsnf4n8aab6e4q0wjhtb89dnuvrjtte

We created a tool for non professional investors to balance their portfolios. We sought to put the power of constraint satisfaction modeling in the hands of casual investors. Users input their favorite stocks and the percentage that they would like to invest in these. Then the Balancer completes the portfolio by minimizing the weighted sum of the correlation across the investment portfolio. It samples from a list of over 100 high performing stocks to balance the users input. The result is a powerful recommendation that provided over 100% returns over 2 years when we backtested our method.

Mentor: Dr. Nicholas Mattei

AnneMarie Donahue, Dorothea Wetmore, Kaitlyn Cooney, Shayna Prinz, Samuel Bruchhaus

Project Link: https://tulane.box.com/s/xv0p4e4emp8cae55gpnpqpnec8zlid0w

The goal of the Tour Planner is to build a tool that will allow musicians and managers to save time planning tours. The tool utilizes two data sources: Songkick, for data on past and future music events, and Spotify, for data on artists and bands. The tool extracts all venues from each city planned on the route. The algorithm then analyzes each venue and creates a profile by analyzing the artists that have played there. By creating this venue profile, the algorithm can match the user to venues in an efficient manner and chart a path through the selected cities.

Mentor: Dr. Ramgopal Mettu

Vasanth Rajasekaran

Project Link: https://tulane.box.com/s/2fq9r896bsexltkizy2hlm0xlzgiqx77

Over-tourism is a growing problem for communities around the world. With travel expected to explode as the COVID pandemic ends, this is the perfect opportunity to develop the means to make sure travel is sustainable from here on out. With smartphones having such a significant influence on how tourists travel, there is much opportunity to leverage smartphones to curb over-tourism. Wayfare is an app that curates a user's day per their interests. The application aims to make planning efficient for people to ensure that little time is wasted and contributed towards the inefficiencies of over-tourism.

Mentor: Dr. Aaron Maus