EAGER: Defending Against Visual Cyberbullying Attacks in Emerging Mobile Social Networks

青少年已经完全接受了社交和交流的社交网络。然而，网络欺凌已经被广泛认为是一个严重的社会问题，特别是对于使用社交网络的青少年。此外，网络欺凌技术也在迅速变化。犯罪者可以利用其移动设备的拍照功能，通过移动社交网络制作和传播受害者的有害照片或视频来欺负他人。虽然网络欺凌的社会和心理方面一直是密集研究的主题，但很少有研究致力于利用计算技术准确和自动地追踪网络欺凌攻击。事实上，自动网络欺凌检测技术的发展仍处于初级阶段。所有关于自动网络欺凌检测的最新研究主要集中在网络欺凌肇事者所写的文本内容(又名。文本网络欺凌)，而在很大程度上忽视了视觉媒体在网络欺凌中的滥用。这项研究表明，要确保对视觉网络欺凌攻击的最强大防御，存在着迫切需要解决的巨大挑战。本项目正在设计和开发一个系统的解决方案，用于自动检测和干预新兴移动社交网络中的视觉网络欺凌攻击。研究人员正在识别与视觉网络欺凌相关的新的人和情景因素，并设计一种用于自动检测视觉网络欺凌的跨特征分类器。该项目还在建立一个自适应的网络欺凌干预系统，以持续监测网络欺凌的情况变化，并为每个相关参与者提供具体的应对策略。提出的方法可以在流行的移动社交网络平台上采用和实施，以防止视觉网络欺凌。该项目设计的系统可以用来遏制年轻人中网络欺凌事件的上升趋势，并减少随之而来的无数有害影响。此外，由于许多针对青少年的其他常见网络威胁现在也必须处理视觉媒体，因此该项目产生的基本成果可以扩大到应对网络欺凌以外的这些威胁。

Adolescents have fully embraced social networks for socializing and communicating. However, cyberbullying has become widely recognized as a serious social problem, especially for adolescents using social networks. Also, cyberbullying techniques change rapidly. Perpetrators can use the camera-capacity of their mobile devices to bully others through making and distributing harmful pictures or videos of their victims via mobile social networks. While the social and psychological aspects of cyberbullying have been the subject of intense study, little research has been dedicated to tracing cyberbullying attacks accurately and automatically using computing technologies. Indeed, the development of automatic cyberbullying detection techniques is still in the nascent stages. All state-of-the-art studies in automatic cyberbullying detection have mainly focused on the content of the text written by the perpetrators of cyberbullying (a.k.a. textual cyberbullying), while largely overlooking the misuse of visual media in cyberbullying. This research reveals that there are great challenges that must urgently be resolved to ensure the most robust defense against visual cyberbullying attacks.This project is designing and developing a systematic solution for automatic detection of and intervention in visual cyberbullying attacks in emerging mobile social networks. The researchers are identifying new person and situational factors associated with visual cyberbullying and designing a cross-feature classifier for automatic visual cyberbullying detection. The project is also building an adaptive cyberbullying intervention system to continuously monitor situation changes in cyberbullying and provide specific response strategies for each associated participant. The proposed methods can be potentially adopted and implemented in popular mobile social network platforms to prevent visual cyberbullying. The system designed by this project could be used to curb the rising trend in instances of cyberbullying among young people and reducing the myriad harmful effects that follow from it. In addition, since many other common cyber threats targeting adolescents must also deal with visual media now, the fundamental results generated by this project could be expanded to cope with those threats beyond cyberbullying.