肉肉传媒

UC-120 Virtual Companionship Chatbot by ; ; ;
Abstract: Loneliness affects about 77% of college students at some point, highlighted by a Gitnuss report. Our project aims to mitigate this by introducing a personalized chatbot that serves as an emotional outlet for students. The application is built on a React Native frontend, employs a DistilGPT-2 language model using the QUAC dataset, and is backed by a Python server. We plan to deploy it on an Azure NC6s_v3 Cloud server, integrating Firebase Real-Time Database for Android and iOS compatibility.
Department: Computer Science
Supervisor: Prof. Sharon Perry
Topics: Mobile Application
| |

UC-105 Model UN Crisis Software by ; ;
Abstract: Utilizing an Agile approach, this project develops a web-based solution, the Model UN Crisis Software, to streamline the management of crisis committees in Model UN conferences. The software is developed using Microsoft Visual Studio and Microsoft SQL Server Management Studio, adhering to the .NET framework and related conventions. It leverages Microsoft Azure SQL Database for back-end data storage and follows the ASP.NET core MVC framework, utilizing powerful .NET tools such as C# and Razor. The developed software provides comprehensive features to reduce the strain of hosting a crisis committee, such as directive and news management, user management, and a messaging system.
Department: Information Technology
Supervisor: Prof. Sharon Perry
Topics: Other
| |

UC-64 Smart Evaluator of Indirect Supplies Vendibility by ; ; ; ;
Abstract: The Smart Evaluator is a web-based software solution that analyzes industrial tools and their vending possibilities. It aims to streamline inventory research for sales teams, reducing manual data gathering and vendibility determination. To begin, users simply upload a basic item inventory spreadsheet, and start the program. From there, the program uses web scraping and ChatGPT to gather key data about the various tools including dimensions, weight, and fragility. Each item is then evaluated based on the collected data, and the optimum storage method is calculated. Once these tasks are performed, the results are stored in the system鈥檚 database for future reference to reduce computation time should a particular item already exist. The overall output of the system is a completed list of items with its necessary data and vendibility options.
Department: Software Engineering or Game Design
Supervisor: Dr. Yan Huang
Project Owners: Vladimir Rusanov and Amir Reza Kashani (Stanley Black and Decker)
Topics: Artificial Intelligence
| |

UR-87 Adversarial Patch Attack in Deep Learning Based Remote Sensing Object Detection Model by
Abstract: Advancements in the field of machine learning have led to object detection systems that can approach or even improve upon human performance. Based on deep learning, these systems play a crucial role in many aspects, and continue to be improved on and see expanded adoption. However, these systems are vulnerable to adversarial attacks that rely on targeted noise to spoof detection. Researchers have applied this concept to increase real world adversarial performance by restricting this noise to a patch that can be placed on new images to disrupt object detection. Previous research has focused on patches applied to person recognition (Thys et al. 2019). We focus on the vulnerabilities inherent in remote sensing systems. We develop an adversarial patch to defeat a modern object detection system, YOLOv8. This work demonstrates that remote sensing can still be defeated by an adversarial patch attack and will inform future efforts to develop model robustness against these attacks.
Department: Computer Science
Supervisors: Dr. Kazi Aminul Islam, Sumaiya Tasneem (Supervising Graduate Assistant)
Topics: Artificial Intelligence
|

UR-89 Performance Analysis of Post-Quantum Computing Key Encapsulation Mechanism Algorithms by ; ;
Abstract: Within the next twenty years or so experts predict that we will have quantum computers which will make certain kinds of encryption that we rely on ineffective and vulnerable to malicious entities. Post quantum computing (PQC) algorithms fill in that security gap that classical encryption algorithms can not. A particular category of PQC algorithms are key exchange mechanism (KEM) algorithm. The goal of these algorithms is to securely generate a shared symmetric key which can be used for encrypting future communication between the hosts. An important use case for these algorithms is in securing the Transport Layer Security protocol (TLS) against quantum adversaries. Due to the widespread use of TLS, it is critical that any new standard use PQC algorithms which are both efficient and secure. To this end we test each of the PQC KEM algorithms provided by oqs-provider library to compare their performance impact on the TLS handshake.
Department: Computer Science
Supervisor: Prof. Sharon Perry
Faculty Mentor: Dr. Manohar Raavi
Topics: Security
|

UR-94 EmoHydra: Multimodal Emotion Classification using Heterogenous Modality Fusion by
Abstract: Affective computing is a field of growing importance, as human society becomes more integrated with machines. Human feelings are both complex and multi-modal, expressed through various methods and nuances in behavior. In this work we introduce EmoHydra, a multi-modal model created through the fusion of three top-level models fine-tuned on text, vision, and speech respectively. Despite heterogenous heads performing well on the unseen data, as well as generalizing well to other benchmarks, logit concatenation proves to be ineffective at predicting Multimodal data, therefore we implement Multi-Head Attention as our fusion mechanism.
Department: Computer Science
Supervisor: Dr. Chen Zhao
Topics: Artificial Intelligence
| |

GPR-103 Personalized Pedagogy through a LLM-based Recommender System by ;
Abstract: The educational domain is undergoing transformation due to the incorporation of Artificial Intelligence (AI), Large Language Models (LLMs), and generative AI technologies, raising the need for educators to integrate cutting-edge technological advancements and methodologies into their teaching approaches. Pedagogical Design Patterns (PDPs) have become prominent for their role in sharing effective educational practices and narrowing the divide between academic research and actual teaching methods. Despite their potential, the lack of widely accessible resources and the scattered nature of publishing outlets pose significant barriers to the broad application of PDPS. To address this issue, we propose the application of large language models to recommend educational strategies based on existing PDPs. Our approach employs a locally sourced knowledge database and the Retrieval Augmented Generation (RAG) framework to formulate context for LLM queries. Preliminary results are encouraging, demonstrating an accuracy rate of 0.83 and a strong relevance of the recommended pedagogical practices to the posed queries. This paper details the initial outcomes of our project, which paves the way for further refinement of the model. The goal is to equip new educators with seasoned insights, to enhance their instructional methods.
Department: Computer Science
Supervisor: Dr. Nasrin Dehbozorgi
Topics: Artificial Intelligence
|

GPR-13 Effect Of Noise And Topologies On Multi-Photon Quantum Protocols by
Abstract: Quantum-augmented networks aim to use quantum phenomena to improve detection and protection against malicious actors in a classical communication network. This may include multiplexing quantum signals into classical fiber optical channels and incorporating purely quantum links alongside classical links in the network. In such hybrid networks, quantum protocols based on single photons become a bottleneck for transmission distances and data speeds, thereby reducing entire network performance. Furthermore, many of the security assumptions of the single-photon protocols do not hold up in practice because of the impossibility of manufacturing single-photon emitters. Multi-photon quantum protocols, on the other hand, are designed to operate under practical assumptions and do not require single photon emitters. As a result, they provide higher levels of security guarantees and longer transmission distances. However, the effect of channel and device noise on multiphoton protocols in terms of security, transmission distances, and bit rates has not been investigated. In this paper, we focus on channel noise and present our observations on the effect of various types of noise on multi-photon protocols. We also investigate the effect of topologies such as ring, star, and torus on the noise characteristics of the multi-photon protocols. Our results show the possible advantages of switching to multi-photon protocols and give insights into the repeater placement and topology choice for quantum-augmented networks.
Department: Computer Science
Supervisors: Dr. Abhishek Parakh, Dr. Mahadevan Subramaniam
Topics: IoT/Cloud/Networking
|

GPR-16 Attention Driven Framework for Detecting Mental Illness Causes from Social Media by ; ;
Abstract: Mental health is a critical aspect of our overall well-being. Mental illness refers to conditions that impact an individual's psychological state, resulting in considerable distress, and limitations in functioning day-to-day tasks. Due to the progress of technology, social media has merged as the platform, for individuals to share their thoughts and emotions. The psychological state of individuals can be accessed with the help of data from these platforms. However, it is challenging for conventional machine learning models to analyze the diverse linguistic contexts of social media data. In this work, we propose a novel attention-driven deep framework to overcome these challenges. Our proposed framework utilizes multi-level (word, sentence, and document) data to identify the causes behind mental illness. The efficacy and effectiveness of our proposed model are shown by extensive evaluation on Reddit data. The insights from this research deepen our understanding of different factors behind mental illness and would aid mental health professionals in formulating effective interventions.
Department: Computer Science
Supervisors: Dr. Md Abdullah Al Hafiz Khan, Dr. Yong Pei
Topics: Artificial Intelligence
|