UROP Openings

Have a UROP opening you would like to submit?

Please fill out the form.

Submit your UROP opening

Techniques for Countering Adversarial Machine Learning


Term:

Summer

Department:

6: Electrical Engineering and Computer Science

Faculty Supervisor:

Martin Rinard

Faculty email:

rinard@mit.edu

Apply by:

May 7, 2020

Contact:

rinard@mit.edu

Project Description

Modern machine learning models (deep neural networks) provide impressive performance on many challenging inference tasks. Research has also shown, however, that they are vulnerable to attacks such as poisoned training data and adversarial inputs that cause neural networks to give unexpected results. This UROP will explore various mechanisms for detecting and/or countering such attacks.

Pre-requisites

Ability to develop and implement machine learning algorithms, including the use of modern machine learning packages such as PyTorch as required.