Poisoning Web-Scale Training Datasets is Practical

This course explores the strategies used by model builders to create large datasets and demonstrates two attacks that exploit these mechanics. The learning outcomes include understanding the risks associated with using web-scale training datasets and learning how to mitigate potential threats. The course teaches skills related to dataset curation, security considerations in deep learning, and threat mitigation techniques. The teaching method involves a presentation format with real-world examples and demonstrations. The intended audience for this course includes data scientists, machine learning engineers, cybersecurity professionals, and anyone working with large-scale datasets in deep learning models.