The advancement of technologies, the promotion of smartphones, and social networking have led to a high tendency among users to spend more time online interacting with each other via the available technologies. This is because they help overcome physical limitations and save time and energy by doing everything online. The rapid growth in this tendency has created the need for extra protection, by creating new rules and policies. However, sometimes users interrupt these rules and policies through unethical behavior. For example, bullying on social media platforms is a type of cyber victimization that can cause serious harm to individuals, leading to suicide. A firm step towards protecting the cyber society from victimization is to detect the topics that trigger the feeling of being a victim. In this paper, the focus is on Twitter, but it can be expanded to other platforms. The proposed method discovers cyber victimization by detecting the type of behavior leading to them being a victim. It consists of a text classification model, that is trained with a collected dataset of the official news since 2000, about suicide, selfharm, and cyberbullying. Results show that LinearSVC performs slightly better with an accuracy of 96%.