Doctor of Philosophy, The Ohio State University, 2013, Computer Science and Engineering

Recent years have witnessed the success of a number of online social networks (OSNs) and explosive increasing of social media. These social networking and social media sites have attracted a significant number of participants that contribute various types of contents on the Internet, which are generally referred as user generated content (UGC). A well designed UGC network can utilize the wisdom of crowds to collect, organize, and vote user contributed content to generate high quality knowledge with a relatively low cost. However, the open environment of UGC system also makes it easy to be polluted and attacked by spammers and malicious users. How users participate in UGC networks, especially how users contribute content and share content with their friends and other users, is fundamental to spam detection and high quality knowledge discovery. In this dissertation, we investigate two important research issues: (1) discovering user content generation patterns in OSNs, focusing on publicly available content (knowledge sharing), and (2) detecting spam in user generated content based on our discovered patterns. With the access to three large OSN user activity logs, including Yahoo! Blogs, Yahoo! Answers, and Yahoo! Del.icio.us, for a duration of up to 4.5 years, we are able to well analyze the patterns of content generation patterns of social network users in detail. Our analysis consistently shows that users' posting behavior in these networks exhibits strong daily and weekly patterns, but the user active time in these OSNs does not follow commonly assumed exponential distributions. We also show that the user posting behavior in these OSNs follows stretched exponential distributions instead of widely accepted power law distributions. Our discovery lays a foundation for user behavior analysis in social networks, and serves as a ground truth for anomaly detection and anti-spam. Applying the user posting behavior distribution pattern, w (open full item for complete abstract)

Committee: Xiaodong Zhang (Advisor); Feng Qin (Committee Member); Ten H. (Steve) Lai (Committee Member) Subjects: Computer Engineering; Computer Science

Master of Science, University of Toledo, 2023, Cyber Security

With the rapid increase of social network spam, it's essential to empower users with the tools to detect the harmful spam effectively. However, existing tools cannot meet the requirements. In this paper, we propose and develop a live detection tool that can detect ham and spam text and images from social networks, this tool will be trained on user collected data (Image and Text) using different classifiers, where text and images are pre-processed and then passed onto the classifier that the user can choose, the user is then able to save the model and load it whenever they want to use a social network, where this tool will show the user a notification alerting them whether the post they are looking at is spam or ham before they even get the chance to read the text or look at the image, thus protecting them from clicking on malicious links that might harm their computer and steal their data. Evaluation results have demonstrated the effectiveness of our tool.

Committee: Weiqing Sun (Committee Chair); Hong Wang (Committee Member); Ahmad Javaid (Committee Member) Subjects: Computer Science

Master of Science, University of Toledo, 2023, Cyber Security

Email communication is a vital component of modern-day communication. However, an increase in spam emails is a significant threat to individuals and organizations which can result in financial and resource losses. Even though the development of effective spam detection mechanisms is essential to safeguarding email security and ensuring safe and secure communication, existing spam email detection methods have some limitations such as the limited capacity to handle the high volume, complexity, and variability of natural language and requiring extensive feature engineering. Moreover, they have the limited ability to remember information from previous time steps, resulting in poor performance, including high false positive rates, low recall, and the inability to detect new types of spam emails. This thesis proposes a novel spam email detection approach that fine-tunes XLNet through supervised training on a labeled dataset of spam and non-spam emails without requiring hand-engineered features. The fine-tuned model is used to predict the class of previously unseen emails. Additionally, the thesis proposes a spam detection model that can handle the high volume, complexity, and variability of natural language in spam emails. The proposed spam email detection model is evaluated on various benchmark datasets, including SpamAssassin, Enron, and Ling-Spam, and its performance is compared with existing models. The model either outperforms or is at least comparable to the state-of-the-art (SOTA) models, achieving an accuracy, the area under the receiver operating characteristic curve (AUC), and F1 scores of 0.9869, 0.9817, and 0.9869 on the SpamAssassin dataset; 0.9892, 0.9893, and 0.9892 on the Enron dataset; 0.9944, 0.9967, and 0.9944 on the Ling-Spam dataset; and 0.9888, 0.9889, and 0.9888 on the combined dataset, respectively. The proposed model's superior performance in detecting spam emails demonstrates its potential to become a key component of email security measure (open full item for complete abstract)

Committee: Jared Oluoch (Committee Chair); Junghwan Kim (Committee Member); Weiqing Sun (Committee Co-Chair) Subjects: Artificial Intelligence; Computer Engineering; Computer Science; Engineering; Experiments; Information Science; Information Technology

Master of Science, University of Toledo, 2016, Engineering (Computer Science)

Online Social Networks (OSNs) have become more and more popular in the whole world recently. People share their personal activities, views, and opinions among different OSNs. Simultaneously, social spam appears more frequently and in various formats throughout popular OSNs. As big data theory receives much more attention, it is expected that OSNs will have more interactions with each other shortly. This would enable a spam link, content or profile attack to easily move from one social network like Twitter to other social networks like Facebook. Therefore, efficient detection of spam has become a significant and popular problem. This paper focuses on spam detection across multiple OSNs by leveraging the knowledge of detecting similar spam within an OSN and using it in different OSNs. We chose Facebook and Twitter for our study targets, considering that they share the most similar features in posts, topics, and user activities, etc. We collected two datasets from them and performed analysis based on our proposed methodology. The results show that detection combined with spam in Facebook show a more than 50% decrease of spam tweets in Twitter, and detection combined with spam of Twitter shows a nearly 71.2% decrease of spam posts in Facebook. This means similar spam of one social network can significantly facilitate spam detection in other social networks. We proposed a new perspective of spam detection in OSNs.

Committee: Weiqing Sun (Committee Chair); Ahmad Javaid (Committee Co-Chair); Hong Wang (Committee Member) Subjects: Computer Science