Dr. Dimitrios Damopoulos and Dr. Georgios Portokalidis
Department of Computer Science
Stevens Institute of Technology
Social media networks have experienced rapid growth and count more than 4.2 billion users. They allow the creation of nodes of individuals, groups and organizations based on trust relationships among them, and facilitate communication between them. Their users store a rich set of personal information, create content based on aspects of their everyday lives, and share their thoughts, concerns, beliefs, emotions and feelings though their personal profiles. At the same time, young children are becoming users of such networks. According to a recent study, 79% of parents who have children at the ages of 2 to 14 years old report that they or their children own some kind of mobile device and access such networks.
Unsurprisingly, these networks have drawn the attention of many aggressors. Over the last year, 95% of social media-using teens who have witnessed cruel behavior on social networking sites, have also been abused from people in their close environment, people they know, (e.g., classmates), or even from strangers that use the same social media platforms. At least 50% of teens have been bullied online, while the 10-20% are bullied on a regular basis according to the iSafe Foundation and the Cyberbullying Research Center. The major threats facing young users of such networks can be categorized into three major types: Cyberbullying, Risky Behaviors and Cyber Predators. The first one, Cyberbullying, relates to the bullying that takes place within a cyber-communication platform (e.g., a chat room) or even in a mobile phone conversation, through SMS, e-mail, etc. Attackers harass their victims by sending hurtful or derogatory messages, threaten, or even force them, into inappropriate behavior, or publish privacy invading and sensitive photos or videos. The second type of attack, Risky Behavior, refers to online communications with strangers that expose sensitive personal information and/or include sexually explicit talk. Last, Cyber Predators mostly refers to sexual predators that target children. According to the Crimes Against Children Research Center, one of five U.S. teenagers, who regularly logs on the Internet, reports that has received an unwanted sexual solicitation via the Web.
Under this prism, we currently try to revolve around a key question: Can we protect children that use social media platforms? For this goal we aim to develop Shield, a framework that can protect children using social media platforms. The Shield architecture targets modern smart devices, which are the main mean of accessing such services, and utilizes the cloud (see Fig. 1 (a)). Shield builds on machine-learning algorithms, allowing us to analyze at run time through text mining a user’s social media activity in an effort to identify attacks like the ones described above. Figure 1 (b), depicts the overall application framework for modern smart devices. Text mining is an analytical approach that focuses on translating the textual data into sentiments and emphasizes on words and expressions generated within a specific context. The process of structuring the textual data involves the use of linguistic techniques and sequential patterns. By utilizing text mining techniques, we focus on discovering the patterns that correspond to the three basic types of threats: Cyberbullying, Risky Behaviors, and Cyber Predators. A key point to this work will be to develop a highly accurate detection system, but also to evaluate it with real case-study scenarios.
Figure 1. Shield: a secure mechanism to protect children’s privacy. (a) Shield Overall Architecture; (b) Shield Application Framework for Smart Devices