Conference Video: Nazli Dereli Adversarial Attacks: A Real Threat to Our Machine Learning Systems
This is Nazlie Dereli’s tech talk at the WiDS Puget Sound Conference 2020. Enjoy!
ABSTRACT:
Autonomous vehicles confusing stop signs with yield signs or authentication systems mistakenly giving access to malicious attackers... Adversarial attacks can trick any state-of-the-art ML system to seriously compromise our security. We'll discuss these attacks and solutions with real-world examples.
Nazli Dereli is an experienced Data Scientist with a demonstrated history of building end-to-end data products. She worked on real-time classification of users and detection of abusive actors in Abuse Prevention team at Ticketmaster for 5 years. In this position, she heavily focused on adaptive abusive behaviors, adversarial attacks and design of evolving ML systems to fight back against bots, brokers, automated systems, ticket scalpers etc. The data products she worked on are being used to protect music fans during ticket sales by Hamilton, Taylor Swift, Ed Sheeran, Twenty One Pilots, Bruce Springsteen and many more. She is a keen learner and published researcher with a M.S. in Computer Science from UC Santa Barbara. Currently exploring new meaningful areas of interest for her data science career while investing in her writing career.