Protecting children and teenagers from cyberbullying

Although their filtering system is able to find malicious chat messages with high accuracy, Two Hat Security was interested in applying machine learning algorithms to automatically detect negative content. To help solve their research challenge, they turned to a Mitacs Accelerate internship with University of Alberta Computer Science PhD candidate Ken Dwyer.

Ken developed an algorithm that can accurately predict the risk of a chat message in a computationally efficient manner. His algorithm was able to identify a number of spelling variants of blacklisted words and evasive patterns that had previously gone undetected by human moderators. Ken’s findings provided Two Hat Security’s moderation team with new tools to increase their efficiency on an ongoing basis, resulting in client satisfaction through improved filtering quality.

Mitacs Accelerate allowed us to explore a new research area that we would not have had the resources or expertise to delve into. As a result, we developed several components that greatly improved our system. The project was a great synergy,” says Two Hat Security’s CEO Chris Priebe.

For Ken, the hands-on experience Mitacs Accelerate offers is second to none. “Mitacs Accelerate provides a wonderful opportunity to get industry experience and take what you learn in school and apply it to real-world problems. Similarly, I can apply the skills I’ve learned in industry to my research in academia.”

After completing his PhD, Ken will continue his role as a full-time employee at Two Hat Security as a lead data scientist.