Detection of Possible Human Trafficking in Twitter

Social networks in general and Twitter, in particular, have become criminal-friendly tools used to contact and deceive their preys and also, for making covert advertising of their illicit activities. In this paper, we collect and process tweets to detect deception related to sex trafficking using predefined criteria as input features to machine learning classifiers. According to the applicable legal law existing in most of our countries, any minor, who is used to participate in a commercial sex act is a trafficking victim. For this reason, in this work, we used the identification of the possible age of the likely victims as one of the essential criteria for the detection of tweets related to human trafficking for sexual exploitation. We tested the validity of predefined features used for the classification of suspected tweets against the rating made by experts. With reasonable precision, we detected tweets possibly related to sex trafficking of underage girls, information that might guide to the police towards antisocial Twitter users and may be useful for law enforcement in the fight against this detestable crime. We used a semi-supervised learning technique with Naîve Bayes and SVM algorithms to classify the tweets as "suspicious" or "not - suspicious" of being related to sex trafficking.