ANONYMIZING PRIVATE PHRASES AND DETECTING DISCLOSURE IN ONLINE SOCIAL NETWORKS
Son Hoang Quoc Nguyen, ソン ホアン コック グエン, Nguyen Son Hoang Quoc · Institutional Repositories DataBase (IRDB) · 2016
Online social networks (OSNs) have become an important part of modern life, and many people use one or more OSNs every day.However, the ease of collecting personal information from OSN messages can make users feel insecure when they access an OSN.Phrases containing personal information, i.e., "private phrases," should thus be identified and anonymized before they are posted on OSNs.Furthermore, if messages containing personal information that are posted only for friends to see are disclosed, the discloser should be identifiable.Most previous research on identifying private phrases has focused on comparing candidate phrases with predefined phrases (such as personal names, locations, or diseases).However, attackers easily recognize and replace the phrases with similar ones (e.g., synonyms, generalizations).Such replacements can be identified by using a co-occurrence metric.In this case, non-private phrases, such as HIV in a non-private message "The human immunodeficiency virus (HIV) is a lentivirus (a subgroup of retrovirus) that causes the acquired immunodeficiency syndrome (AIDS)," might also be identified.We have developed an algorithm for determining whether an OSN message is a private message or a non-private one.The F-score was 92% for 3000 messages, showing that it works well for identifying private messages.The algorithm identifies