HOMIEZ is a web application to provide users with real-time visibility into property price changes, ensuring they stay informed about market fluctuations. By eliminating unnecessary distractions such as advertisers and focusing solely on price tracking, the platform empowers users to make data-driven decisions without relying on delayed updates or extensive searching.

Blacksburg Trip Advisor is a web application specifically marketing to primarily visiting high school students, incoming undergrads, and incoming graduate students, but applicable to all visitors. It allows filtering based on time/date/location/keywords to search for current events, relevant landmarks, and restaurants; the goal is to provide a consolidated site for all travel information as well as a tool that produces a potential schedule for guest-based on their feedback and travel goals. The website combines information from both on-campus and off-campus which helps people to find anything without needing to search on multiple websites.

​This project explores machine learning-based detection of OS Command Injection (CWE-78) vulnerabilities, focusing on precisely locating security flaws within source code. By integrating static analysis and deep learning techniques, the study enhances vulnerability detection by mapping root cause and manifestation points in program dependency graphs. Using Graph Neural Networks (GNNs) and Structure2Vec models, the approach improves classification accuracy and localization of vulnerabilities. The findings contribute to advancing software security by refining automated detection methods, reducing false positives, and aiding developers in efficient vulnerability remediation.

This lab showcased the real-world impact of XSS vulnerabilities and their potential risks in web security. Conducted an XSS exploitation demonstration using the Elgg web application in an Ubuntu VM. Successfully launched and propagated an XSS worm, compromising session cookies and leveraging HTTP GET and POST requests to spread the attack to profile visitors and friends.

In this study, a comprehensive approach to fake review detection, leveraging the RoBERTa model is proposed. Our approach combines the extraction and processing of the intrinsic features of fake reviews and RoBERTa transformers to improve the model’s ability to distinguish between authentic and fraudulent reviews. The feature-enhanced RoBERTa, FeaRoBERTa, uses a vectorized feature processor to incorporate non-sequential features and RoBERTa to handle tokenized sequences. The combined features are then fused and passed through a linear classifier for binary classification. 

This project examines the effectiveness of Instagram influencers in marketing by analyzing a dataset of 33,935 influencers and over 10 million posts. Using statistical analysis, sentiment analysis, and engagement metrics (likes, comments, followers, and followees), the study identifies key factors that contribute to an influencer’s success. Regression models and content analysis were applied to explore correlations between influencer attributes and audience engagement, providing insights into category-based influence, audience sentiment, and marketing strategies. The findings offer valuable guidance for brands in selecting the right influencers to optimize their social media marketing campaigns.