Ketan Jadhav

Publication Date

Spring 2024

Degree Type

Master's Project

Degree Name

Master of Science in Computer Science (MSCS)

Department

Computer Science

First Advisor

Katerina Potika

Second Advisor

Philip Heller

Third Advisor

William Andreopoulos

Keywords

Bot identification, Natural language processing, node classification, embedding techniques, twitter, graph neural networks, TwiBot22.

Abstract

Social media is a key resource in modern human communication as well as for information. Ease of access and global reach is a primary factor to the popularity of several social media platforms like Twitter. Social Media Bots are automated programs which are developed for social engagement. These bots, however, are being used with malicious intent as well, to spread fake news and manipulate the masses. Identification of social media bot accounts has become crucial since social media has become one of the primary sources of news and information for a lot of people. This project aims to propose Multirelation Bot Detection Graph Neural Network methods (MultiBotGNNs)to solve the bot detection as a node classification problem. We incorporate ideas from existing techniques for bot identification and use Natural Language Processing, and Deep Learning. This project is developing an approach to identify social media bots by using the user characteristics, their social media activities, the user account bio processed with NLP along with the social network graphs of different user relations, as follow, follower, and like. We experimented with the TwiBot22, which is a big dataset of bots. Our focus was to consider the structures of the relations for the prediction, and ignoring the tweet contents and show that we can good results without the extra process time for the tweets.

Download

Available for download on Sunday, May 25, 2025

Share

COinS