Publication Date

Spring 2024

Degree Type

Master's Project

Degree Name

Master of Science in Computer Science (MSCS)

Department

Computer Science

First Advisor

Robert K Chun

Second Advisor

Ching-Seh (Mike) Wu

Third Advisor

Chandra Pavan Reddy Chada

Keywords

Cyber Threats, Data Breaches, Fraud Detection, Malicious Links, Natural Language Processing, Phishing Attacks, Social Engineering, Transformers, URL Analytics

Abstract

In short, the incidence of phishing - the illegal act of people pretending to be well-known companies to secure personal information - has skyrocketed in the past few years. In 2022 alone, 300,000 consumers in the United States were captured by scammers using phishing techniques, losing in all over $50 million. In the span of two weeks, over 510 million attempts occurred in a variety of sectors, particularly instant messaging platforms, package delivery businesses, and digital currency trading. Since most businesses have recognized that they are prone to these exposure cases, there has been a sixty percent increase in businesses at the same time as the preceding year. Mislead connections have been a big issue, and verification screens have been taken over. Traditional methods like TF-IDF and logistic regression samples frequently mistake benign transactions for dangerous threats. The solution avoids this through strong NLP algorithms to gather further specialization via Uniform Resource Locators. It leverages DL algorithms like LSTM networks with Transformers to identify the complex target's structure. Abnormal Uniform Resource Location structures shall be identified by anomaly recognition along with system's adaptive learning that shall adapt constantly to identify fraud. Therefore, specifically, this innovative technology is unique in that it is embedded with analytics on user behavior, broadening the detecting capability of defenses against the wider variety of phishing.

Download

Available for download on Friday, May 23, 2025

Share

COinS