Adversarial Attacks on AI Models

This project is significant for security in models, in particular in the financial markets.  You would be asked to try some basic adversarial attacks under some specified constraints on some time-series regression models to test a basic hypothesis, and depending on how that goes, do some follow-up work. The goal would be to find ways of modifying the data such that a time-series model fails to perform. This is an experimental project and is in the early stages, so what precisely should be done here is still open, but there is also lots of opportunity to shape the project at this stage.