Researcher Ayat Awad Najjar, a student in the Master’s Program in Cyber Security has defended her thesis titled "Detecting Written Documents by ChatGPT for the Cybersecurity Domain Using Machine Learning".
This research focuses on creating a robust model for detecting texts produced by large language models popular in the field of natural language processing, based on the assumption that machine learning technology can detect machine learning technology. The first study explores the field of cyber security and highlights the potential risks associated with using AI-generated texts in malicious ways.
The study delves into creating interpretive AI techniques designed to differentiate human-written texts produced by large natural language transformers like Chatgpt by training and evaluating different machine learning and deep learning algorithms. In addition, the second study explores the field of academic integrity, emphasizing the obstacles faced by students who rely on large language models to complete their academic assignments. Because traditional citation detection techniques compare texts produced by large language models with historical information on the Internet, they often fail to recognize texts produced by these models. The researcher also addresses the potential impact of excessive dependence on artificial intelligence tools, which results in a generation of students who lack important expression skills.
This study trained and evaluated different machine learning and deep learning algorithms on the databases that were created for the two aforementioned studies. Through research, these documents were discovered using machine learning, deep learning, and interpretive artificial intelligence techniques. Also, the model used performed better than the commercial AI detection tool GPTZero when comparing accuracy.
The thesis was supervised by Dr. Omar Darwish and co-supervised by Dr. Hudayfa Al Ashqar. The committee of examiners included Dr. Osama Mansour and Dr. Islam Amro.