This paper presents some preliminary work on detecting sex offenders who might register in social networks under assumed identities using data mining, string matching, and weight-biased scoring.