Privacy Protection Protocol in Social Networks Based on Sexual Predators Detection
Zeineb Dhouioui, Jalel Akaichi · 2016
Social networks offer great opportunities for communication and no one can deny the benefits of these sites; however, we cannot ignore the privacy defects such as fraudsters, spam, hackers and sexual predators. This paper presents a privacy protection protocol in order to protect users from risks that may be encountered in their social networks activities. In particular, we focus on the sexual predators detection. For this purpose, we use text mining tools to classify doubtful conversations based on lexical and behavioral features extraction. Moreover, we are able to flag potential predators, by computing the predatorhood score according to the sum of features weights. Different experiments have been carried out based on comparative study between two machine learning algorithms: support vector machines (SVM) and NaÃŕve Bayes (NB). The results are very promising.