Research Project Title:
Exemplary Data Anchor Points as an Effective Neural Network Defense Mechanism
abstract:Neural networks have been the focus of modern machine learning research due to their established success on a broad range of tasks, from image classification to machine translation. However, recent work has brought into question the robustness of these models to attacks in which adversaries craft data examples deliberately designed to be misclassified by state-of-the-art networks. The goal of this project is to enhance the robustness of neural network classifiers using methods that are more efficient and effective than the current approach of adversarially retraining the network. In particular, we will implement and evaluate a novel data-driven defense mechanism in which exemplary anchor points for each output class are used to impose additional regularization on the classification task.
After taking graduate classes in machine learning and security last year, I have a growing interest in tackling problems at the intersection of these two disciplines, which is what drew me to this SuperUROP. I hope to not only expand my technical knowledge and mathematical maturity, but also to improve my communication skills and get a better understanding of the entire research pipeline, from formulating an idea to presenting my work.