This article discusses a simple method to detect whether an online conversation is grooming or not. This issue is of importance by considering that online child grooming is the most reported crime. The current method identifies seventeen characteristics that commonly exists on the grooming conversation and their existence on the conversation determines its status: grooming or not. This study found that a textual conversation containing eleven or more characteristics is likely to be grooming. The analysis of 100 grooming conversations from www.perverted-justice.com and 50 non-grooming conversation from www.literatorika.com reveals that the current simple method can achieve 97\% of the level of accuracy. This result is a slightly better than that of the previous logistic-regression-based method. The present method requires much lower computational resource.
Fergyanto E Gunawan Dr Eng and Livia Ashianti
Published at :