Research has shown that over 40% of Taiwanese students engaged in cyberbullying relationships in 2020. School librarians play a key role in educating students about information literacy and cyber citizenship, making it important for them to develop cyberbullying awareness and prevention policies. This study proposes a cyberbully detection model, with a particular focus on body shaming. Therefore, body shaming has become increasingly ubiquitous with the rise of social media. We collected over 37 million comments from Taiwan’s largest internet forum, PTT Gossiping Board via web crawling. To distinguish specific terms used in body shaming, we tokenize words that are directly related to “fat” and “thin” through the CKIP tagger. Since online users would often use metaphors to describe body shapes, this study also develops a user-defined dictionary to cope with this problem. We identified 7 characters from the body shaming comments using the Objectified Body Conscious Scale-Youth (OBC-Y) scale. The characters include gender, pubertal development, sexual harassment, appearance-related teasing, flaming, shaming, and appearance control belief. We then applied machine learning algorithms to detect these body shaming features. The result showed that the decision tree algorithm had a precision rate of 95.48%, which surpassed the precision rate of the Naïve Bayes model. Additionally, the most weighted feature was "gender," specifically words related to the LGBT community.