Predicting Malicious Behavior: Tools and Techniques for Ensuring Global Security

Chapter 15

Predictive Behavioral Modeling: Automated Tools of the Trade

Throughout this book, I have mentioned automated behavior analysis (AuBA) and its associated tools: ThemeMate and AutoAnalyzer. Using these two tools properly with an adequate text corpus describing past incidents enables the construction of behavioral models that can predict future behavior in a valid and reliable manner. Although I developed a manual process first, AuBA tools can result in the construction of a predictive model 400 times faster than the manual process. Even the manual process had to have some components constructed in software, such as the pattern classification component. However, as it exists today, AuBA has many features that require automation to be totally effective. For example, the ThemeMate tool that processes text examples of past events and identifies antecedents and creates the data array of advanced pattern classification does not exist in the manual method. AutoAnalyzer, which automatically constructs and validates the predictive engine, is exclusive to AuBA and is not in the manual method.

Because of the AuBA components, the ability to identify relevant antecedents in past text examples has increased significantly. In the past, using the manual method, we would tend to identify fewer than 100 antecedent candidates. We limit ThemeMate at 500. In addition to the many new features that AuBA has over our manual process and the automation is a reduction of bias. As the previous chapter ...