Keywords

1 Background

1.1 AiBA

This project is conducted as part of the AiBA (Author Input Behavioral Analysis) project by the Norwegian Biometry Lab at NTNU Gjøvik. AiBA aims at identifying fake profiles in chat applications using behavioral biometrics, in particular for protecting children from sexual predators that find their victims online. For example, behavioral biometrics can reveal an adult pretending to be a teenager in order to groom children online. The algorithms analyse word usage, writing rhythm, linguistics and media input to differentiate between adults and children as well as between males and females. The system is trained with data from convicted abuse cases and chat data acquired from children using a chat prototype. It is envisioned that the algorithms will either be built into platforms and applications used by children, such as Roblox, Snapchat and Instagram, or will act as a standalone application that retrieves data from the chats. Alerting the users that such a security measurement is in place will contribute to deterring those with dishonest intentions, since there is a higher chance of being disclosed.

1.2 Online Grooming

Grooming describes behavioral patterns that are applied by sexual predators in preparation to sexually abuse a child. Colton et al. [4] describe grooming as a complex process in which the predator gains access to the chosen victim so that abuse can be initiated and maintained without being disclosed over time. However, literature shows no consensus over the specific methods that are employed by offenders [2] and empirical research has found that it is hard to recognize grooming behavior prior to abuse since it is not always clearly distinguishable from normal adult-child interactions [16]. Nevertheless, Winters et al.  [16] describe that almost half of all convicted child abuse cases in the US have been preceded by what was identified as grooming behavior. Lanning [10] developed an accurate description of the process and emphasizes that grooming is a non-violent practice that aims at sexual victimization and control without using threats or physical force, as these techniques are more likely to result in cooperation from the victim’s side. Lanning [9] describes the process of grooming as incremental stages, starting with identification of targets based on certain factors, followed by gathering information about the child’s interests and vulnerabilities. The offender then gains access to the victim through a variety of channels, for example through clubs, sports or online. The victim is controlled by the offender by a combination of strategies such as filling emotional or physical needs, bonding, sympathy or peer pressure. The internet and its infinite number of chat platforms has opened up a new channel for child offenders to gain access to victims and provides the offenders with anonymity and a wide reach to locate victims. Goals of online sexual grooming include cyber-sexual activity, access to child pornography or arranging in-person meetings [8].

1.3 Risk Communication

Risk communication describes a set of communication measures that prepares the audience for informed decision making before, during and after an event [5]. Risk communication originates from the public health sector where it is used to communicate and educate about health risks and epidemics. The approach has also been adapted to communicate risks of natural disasters such as floods and fires, but it can also be practiced on a smaller scale, for example in the practice of warning messages and labels [11]. Lundgren et al. [11] point out that there are a vast number of approaches to risk communication that have evolved from various disciplines and that the approach should be chosen according to the context and the audience. The risk communication process usually starts with a risk assessment that evaluates who will be harmed by the hazard and what effects can be expected, as well as how long these effects last. A strategy is developed how awareness can be raised about the subject and how the audience can be educated. Information about how to take action in case of an event needs to be communicated in a clear and understandable manner. Also, a risk communication strategy should be disseminated through channels that are frequented by the target audience [11]. In addition, it is recommended to design messages according to the audiences’ knowledge, interests and values [3]. Lundgren et al. [11] emphasize that the more the communicators understand the perspective of the target audience(s), the better they can choose appropriate communication methods and thus implement successful risk communication. This shows the prominent role that user research should play in risk communication and justifies the user-centred approach taken in this project.

2 Methodology

2.1 Focus Groups with Children in 7th Grade

With the goal of acquiring profound insights into the perception and knowledge about online sexual predators and grooming strategies, focus groups were designed to investigate children’s chat behavior, how they react to certain messages in chats and to what extent they are aware of dangers. Two seventh grade classes with a total number of 35 pupils participated in two project days about “Chat Security”. To investigate the impact of education on the subject, two focus groups were conducted before a presentation on chat security and grooming strategies was given, and four groups after, resulting in six focus groups being conducted at the school, each consisting of 5–6 children of 12 to 13 years. During preparation, it was taken into consideration that working with children results in certain challenges such as differences in children’s development due to their culture and social environment [1], ensuring the children’s engagement and attention [7] as well as their safety and comfort [12]. It was investigated how they think they can trust someone online and where they draw the line between appropriate and inappropriate messages. The focus groups concluded with an assessment of the concept behind the AiBA project.

The focus groups were divided into three sections. In the opening part, a brainstorming session was conducted, first on what the participants use the Internet for, and then more specifically what applications they use for chatting with other people. This data can be used to establish the most effective locations for placing information about online predators and also potential customers for the AiBA application. To challenge the children’s creativity, they were then asked to come up with a username for a fictional chat. Since usernames are one of the targeting criteria of sexual predators [17], the data will be helpful to determine how children choose them. After presenting the usernames, the moderator challenged the participants by exchanging her username with one of the children’s. They were asked to come up with ideas how to find out who really is behind a nickname and how to trust someone online. The second part of the focus group consisted of a made-up chat conversation that followed a simplified grooming process. The children were challenged to come up with replies or reactions to the messages which got increasingly inappropriate. After each message, the children were asked if they thought this message was written by someone of their age, an adult or if it could be both. This data will give insights about how children behave in chats and where they draw the line between appropriate and inappropriate conversations. As the last part, the AiBA project was summarised in a simple and understandable way. The children were then asked to rate how much they liked the idea of such a system by assigning a number (1 being worst and 10 being best) on a sheet of paper. They were also asked to write down a reason for their decision. After each focus group, some notes were taken on the children’s overall behavior and reactions, as well as their body language and other remarks.

3 Results

The collected data was mainly of qualitative nature, showing opinions, experiences and reactions of the children. This data was analysed by applying a mix of design methods such as thematic evaluation and affinity diagramming. Exploratory statistical analyses was also used. Special attention was given to understand differences between the groups who have not received education about the subject and those who have. Already on location during the project days, it was possible to notice some immediate results. It was clear to see that there are immense individual differences in how the pupils reacted to the prompted chat messages, some being very honest, others extremely careful or evading through humour and irony. On the second day - after the presentation on chat security and online grooming - it was very striking to observe how much more careful and suspicious the children were acting. Their behavior changed and they were more direct in differentiating between appropriate and inappropriate messages. Also, the terms “sexual predator”, “kidnapper” and “pedophile” were mentioned regularly, while this had not occurred on the previous day.

After an in-depth analysis, one could see that educating children about chat security and grooming can show immediate, short-term results. The children became more aware of the risk and were discussing about sexual predators several times during the focus groups if they had been educated before. The focus groups showed the popularity of using Snapchat among children of 12 and 13 years as well as their strong interest in playing online games and consuming media content. These platforms can be important partners to spread education on online grooming and could also be potential customers for the AiBA system. The children mostly chat with friends, but it is not uncommon to chat with strangers when it is about a common interest, such as gaming. The children felt confident that they could detect an adult who tries to pretend to be a teenager through the language that is used and also the content of the conversation. They rated pictures or video calls as the most reliable option to find out about someone’s real identity. When creating usernames, the children unintentionally gave away personal information such as their name and age, regardless whether they had attended the presentation on chat security or not. In a fictional chat environment, those children who received education on grooming before were more self-assured to say “no”, acted more confident and suspected dishonest behavior sooner than children without previous education on the topic. The overall rating of the AiBA system was more positive after the presentation, showing that the children saw more value in it than before.

The insights were further evaluated with a set of gamestorming methods as described by Gray [6], resulting in a risk assessment. The reason for a need of education are the high level of Internet access and chat usage at an increasingly younger age, meanwhile predators are using chats as a means for easy access to victims. As estimated by a recent EU study [14], children aged 12–14 years in the EU spend around 192 min online each day. Children are particularly at risk once they start conversing with strangers and give away private information, either unknowingly (i.e. through usernames), due to risk-seeking behavior (doing something “forbidden”) or simply because they are not aware of the danger and engage in seemingly harmless conversations. The European Online Grooming Report [15] describes that children that are perceived by predators as particularly vulnerable are more likely to be targeted, as well as risk-taking children who are outgoing and confident online - similar to the children in the focus groups who reacted to the chat messages with humour or by being offensive. However, while it is important to take these risk factors into account, children at risk should not be stigmatised or reduced to a homogeneous group as both individual behavior and targeting tactics of online predators vary greatly. Challenges that are faced when tackling the topic are the general unawareness of the children who are active in chats and also the parent’s lack of knowledge about the subject. It is sometimes treated as a taboo topic that evokes feelings of shame, it can result in victim-blaming and is seen with hindsight-bias, meaning that people claim to have “seen it coming” only after an incident, or the risk is being ignored as it seems unlikely to happen to oneself. Since children want to maintain a certain level of autonomy and privacy, a lack of parental supervision can increase the risk, as well as a lack of security measures in many chat applications. However, if the risk is not addressed, it means that ultimately more children will fall victim to sexual abuse, either through physical meetings with the predators or through online activities. Children will be exposed to sexual content and might share private pictures and information. This can result in the children being bullied, threatened or blackmailed which in turn results in psychological damage and a dependency on the relationship with the predator. In real life, consequences can be seen in a problematic relationship with the parents and peers, since the online relationship gains importance. Therefore, the goal of the risk communication is mainly preventive, aiming at children who engage in chat environments, whose safety and security as well as privacy and autonomy should be secured. Educating them about grooming will increase their caution on the web, enable them to protect themselves and give them the courage to report incidents.

As children spent a vast amount of time on the Internet, spreading information directly in the chat environments, on social media or gaming platforms is believed to be an appropriate choice of media. Furthermore, postings on social media can be used to target specific user groups that are at a higher risk. The children should be able to relate to the messengers that convey the content, therefore it is important that the messenger is close to the peer group and that the wording appeals to them. This is believed to increase the children’s interest in the content and work against barriers such as ignorance and shame. Another effective way to reach the children is to conduct seminar days at schools similar to those conducted during the user research. This can provide the children with valuable background knowledge of Internet security and online behavior that can encourage them to rethink and adjust their own behavior. Building on this, the messages for precautionary advocacy can be designed, taking into account the guidelines for warning design by Sandman [13]. Thus, the messages need to be on point, raise interest, appeal to emotions and offer choice. Short, concise and understandable messages ensure that the information is interpreted in the intended way. Additionally, messages need to be interesting and appealing to the audience so that they are heeded in the first place. In combination with appealing to emotions (especially fear in proportion to the risk), chances are high that people will pay attention. Offering choices of actions to take, preferably in easy and small steps, encourages necessary changes in behavior. The main questions that need to be answered at this stage are: How does online grooming happen? What are the consequences? How can I protect myself?

The immediate response phase to a risk is triggered when the AiBA system detects suspicious behavior that indicates predatory behavior. A warning has to be sent so that the child and the parents can decide how to respond to the threat. For creating these warnings, design principles can be applied regarding the structure, understandability, noticeability and context of the messages [18]. The goal of the warning message is to make the user aware that the chat partner is showing signs of predatory behavior or grooming. The message needs to attract the users attention, so it needs to be prominently displayed and designed. Timing is crucial as sending a warning too late can result in harm to the child, while sending it too early can result in unfounded fear and false accusations. In order to justify the message and to ensure understandability, it needs to be explained how the system arrived at this conclusion, for instance by listing the suspicious characteristics of the conversation. The risk and its consequences should be explained briefly to mark the urgency of the situation and the need to take action. The warning needs to present choices and recommendations for subsequent actions in easy steps. For instance, a checklist can help the user to identify grooming behavior and if the chat partner should be blocked and reported. Since AiBA will not be infallible, the final decision to take action lies with the user.

After the immediate risk has been averted, the audience should be given options to recover and react to the incident. The focus of the recovery phase lies on providing support and counteracting apathy. The information should be sent directly to the audience so that they immediately have access to it after AiBA warned them about predatory behavior. Additionally, the information should be available on demand so that it can be reviewed at all times. Crucial points to consider are providing emotional support, indicating institutions where one can receive additional help and encouraging that grooming should be reported. Again, the information should encourage young people to draw lines and stand up for their own well-being.

4 Discussion

The short term results of the focus groups show that education about online security and grooming does seem to influence behavior and makes children more aware of risks on the web and thus more careful. The extensive use of the web and particularly chat environments by young children shows the importance of this project. Also, the high interest that the children have shown towards the topic gives hope that education measures will be received eagerly and the seriousness of this is understood very well. However, further assessment is needed to establish long term results. The importance of education on online grooming for raising awareness and preventing abuse is backed up by previous research [15]. Establishing information needs and extracting the key messages for each stage of risk communication allows for differentiated information distribution that provides the audience with relevant information when it is needed.

Work in Progress: Additional focus groups with children are planned to strengthen the results and gain additional insights. Furthermore, the communication strategy will be supplemented by a similar approach that provides information for parents, since they serve as role models and close confidants of the children. Future work will consist of refining the strategy and creating a set of tangible messages and information materials that will be evaluated and adapted based on user tests. The completed strategy will be presented in a master’s thesis at NTNU and is hoped to be implemented with the completion of the AiBA system, providing children and parents with both education and practical means to protect themselves and their loved ones on the web. The interdisciplinary nature of this project is meant to encourage people working in design, risk communication and computer security to cooperate. It shows that user-centered design approach can be applied to a variety of fields outside the traditional design context. A multi-faceted risk such as online grooming needs to be managed as a collaborative approach in order to promote prevention and minimise the degree of harm.