Multimodal Cyberbullying Detection on Social Media: Review and Challenges

Cyberbullying is a prevalent issue on social media platforms, where people employ various digital means to harass and intimidate others, particularly affecting young individuals. According to a study by the Cyberbullying Research Center, 34% of students aged 12–17 have experienced cyberbullying. Unfortunately, cyberbullying is frequently underreported, with only 10% of affected students reporting it to adults. To tackle this problem, machine learning-based automated detection systems have been developed. However, single modality detection systems that focus on analyzing only one type of data, like text or images, may have limited accuracy due to the complex and variable nature of cyberbullying behavior. This paper explores the importance of multimodal cyberbullying detection on social media and the challenges involved in creating effective detection systems. It examines existing literature on multimodal cyberbullying detection, emphasizing the advantages of analyzing multiple modalities. The paper also addresses the ethical considerations associated with analyzing personal data to identify cyberbullying behavior.