A hybrid approach to combat email-based cyberstalking

Email is one of the most popular Internet ap- plications which enables individuals and organisations alike to communicate and work effectively. However, email has also been used by criminals as a means to commit cybercrimes such as phishing, spamming, cyberbullying and cyberstalking. Cyberstalking is a relatively new surfacing cybercrime, which recently has been recognised as a serious social and worldwide problem. Combating email-based cyberstalking is a challenging task that involves two crucial steps: a robust method for filtering and detecting cyberstalking emails and documenting evidence for identifying cyberstalkers as a prevention and deterrence measure. In this paper, we discuss a hybrid approach that applies machine learning to detect, filter and file evidence. To this end we present a new robust feature selection approach to select informative features, aiming to improve the performance of machine learning within this task.