Automatically Identifying Online Grooming Chats Using CNN-based Feature Extraction

Automatically Identifying Online Grooming Chats Using CNN-based Feature Extraction Svenja Preuß and Tabea Bayha and Luna Pia Bley and Vivien Dehne Alessa Jordan and Sophie Reimann and Fina Roberto and Josephine Romy Zahm and Hanna Siewerts and Dirk Labudde and Michael Spranger University of Applied Sciences Mittweida Mittweida, Germany spranger@hs-mittweida.de Abstract With the increasing importance of social me- dia in everyone’s life, the risk of its misuse by criminals is also increasing. In particular children are at risk of becoming victims of on- line related crime, especially sexual abuse. For example, sexual predators use online groom- ing to gain the trust of children and young adults. In this paper, a two-step approach using a CNN to identify sexual predators in social networks is proposed. For the identification of a sexual predator profile an F0.5 score of 0.79 and an F2 score of 0.98 were obtained. The score was lower for the identification of specific line which initialized the grooming pro- cess (F2 = 0.61). 1 Introduction The importance of social networks in today’s so- ciety is constantly growing. More and more chil- dren and young people are turning to digital forms of communication. Studies from Germany show that 71% of children between the ages of 6 and 13 actively use the Internet, and the trend is rising (Feierabend et al., 2020b). The situation is similar for young people between the ages of 12 and 19 (Feierabend et al., 2020a). In one study, 97% of the teenagers surveyed said they used the Internet every day or at least several times a week (Feier- abend et al., 2020a). Those developments provide new opportunities for sex predators to gain access to minors, for example, through online grooming. The Austrian Federal Criminal Police Office (Bundeskriminalamt, 2015) defines online groom- ing as the targeting of children and young people on the Internet with the aim of establishing sexual rela- tionships. It is a special form of sexual harassment that can lead to physical and sexual abuse. The contact is initiated via the Internet, for example via social media or online video games. In child online grooming an adult predator uses means of online communication in order to gain access to and trust from a minor in order use the minor for sexual purposes (Wachs et al., 2012). In many countries, cyber grooming is legally considered a criminal offense. In the U.S., for example, 18 U.S. Code § 2422 criminalizes on- line grooming. In 2011, the European Parlia- ment passed Directive 2011/92/EU, which obliges member states to enact corresponding legal reg- ulations, including on criminal prosecution. In Germany the criminal law aspect was regulated in § 176/IV StGB. In an effort to contain such sexual offenses software to identify potential predators is devised (Inches and Crestani, 2012). That kind of Soft- ware is supposed to be a preventive measure whose forensic/criminalistic benefit lies in assisting the day-to-day police work and even possibly prevent- ing sexual offenses from happening. The goal is to reduce the expenditure of …

Content truncated to 500 words