The rise of social media has enabled online aggression practices such as cyberbullying, harassment, and hate speech to reach an unprecedented scale. Some aggressors select their targets and coordinate on polarized online communities to organize attacks against their victims, inundating them with hateful or disturbing messages, videos, and images. These attacks can cause serious harm to their victims, forcing them to leave social media sites or even to contemplate self-harm. Despite the threat posed by online aggression, the problem has so far not received much attention by the computer security research community. Developing quantitative methods able to identify and mitigate such attacks is however of paramount importance to provide a safe online experience to all Internet users. This project aims to develop tools able to identify online aggression attacks in real time, allowing online services to take the appropriate countermeasures.
In this project, the PI aims to achieve four research objectives. First, by leveraging annotation from crowdsourcing workers, this project aims to develop effective corroborating evidence of aggression incidents on social media. De-identified datasets will be released publicly and will help the research community at large to better understand the problem. Second, the PI will develop techniques based on machine learning to identify online accounts that partake in online aggression, and to automatically flag the hateful content that they post, allowing online services to quickly react to such attacks. Third, the project will develop predictive models to establish the likelihood for content that is posted online to receive hate in the future; online services will be able to use these models to proactively allocate moderation resources towards content that is considered at risk. Finally, the PI will investigate the advantages and disadvantages of different mitigation approaches for this problem, from suspending offending online accounts to disabling comments for particularly risky content. This project's educational activities will include an interdisciplinary module targeted at college students from non-technical backgrounds and a tutorial designed to provide high school students with a better understanding of cyberbullying attacks.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
