Applying a machine learning model to forecast the risks to children's online privacy and security

Children are using digital gadgets at an increasing rate around the world, and this is accompanied by an increase in online privacy and security risks for them. Earlier studies found that children, parents, and teachers can work together to improve children's privacy and security when using digital devices. However, it was also discovered that safety measures can be applied more effectively if it can be predicted earlier which children are at risk and who are not. This research proposes a machine learning model based on 12 data points from children and their parents to predict children at risk of online abuse, cyberbullying, sexting, and stress. The model was created using the interview findings of 42 parents and 30 children as well as separate survey data from 311 parents and 383 children. According to the data study, children suffer the most from abuse, bullying, sexting, and severe mental stress among the various forms of privacy and security violations in the online world and that is why the prediction was done on these 4 situations. In this study, the Random Forest algorithm was used to develop the machine learning model and it was found that the model can somehow predict the privacy and safety of children while using digital devices. Our study contributes to the literature on Computer Supported Cooperative Work (CSCW) research that aims to develop cognitive and conceptual models to make the digital environment safer for children and adolescents.