11:00am - 11:30am

Dr. Tahir Ekin Gregg

McCoy College of Business, Texas State University

Title: Adversarial forecasting

Abstract: Statistical and machine learning methods used for forecasting typically assume clean and legitimate data streams. However, adversaries may attempt to influence data and alter forecasts, which in turn may impact decisions. This talk presents an adversarial risk analysis-based framework that allows incomplete information and adversarial perturbations. We solve the adversary’s decision problem where he manipulates batch data. This research highlights the weaknesses of forecasting models under adversarial activities. It motivates the need to improve the security of existing decision frameworks and to proactively address the potential manipulations. We illustrate the proposed method using hidden Markov and ARIMA models, present an overview of computational algorithms, and discuss potential applications.