Texas State University,
Keywords: adversarial machine learning, counter AI, Bayesian, adversarial risk analysis
Summary:Statistical and machine learning methods typically assume clean and legitimate data streams. However, adversaries may attempt to influence data in order to impact decisions in a way that would favor their own objectives. The information about the adversaries’ goals and capabilities are not known by the artificial intelligence methods in advance. Probabilistic methods such as Bayesian approaches could be used to incorporate such uncertainty within these methods. In this presentation, we focus on examples from adversarial time series and outlier detection methods, and illustrate the use of adversarial risk analysis within adversarial probabilistic AI methods.