This document provides an overview of adversarial modeling techniques for fraud detection. It discusses using machine learning, graph theory, and text analytics together in an agile process. Unsupervised learning and graph networks are important for discovering fraud patterns. Text analysis can link similar documents and be incorporated into models. The problem requires cross-functional teams and deploying solutions iteratively to rapidly respond to adversaries' changing behaviors. Rather than a single approach, an ensemble of data models, tools and techniques works best.