Elite Data Analysis For Security IV: Statistical Modeling

By Todd LaRocca

statistical-modeling-methods-for-security-data-analysis.jpgThe difference between mitigating and managing security threats at your federal agency boils down to one major factor: your ability to predict them. 

Whether you’re monitoring potential adversaries or predicting insider threat behaviors, statistical modeling is a critical component to agency security because the more incidents you’re able to predict and prevent, the fewer major security incidents you’ll face. 

In this fourth post of a five-part series, we take a deeper dive into the data analysis tools created from statistical modeling – and how to use those tools to proactively address security threats at your agency. 

In Part I, we discussed robust data collection tactics for stronger security decision-making; in Part II, we examined how to avoid common pitfalls when working with data dashboards; and in Part III, we drilled into the importance of data quality for security decisions.

If you want to predict and proactively address security threats at your intelligence or defense agency, you must learn to leverage your past data for future predictive modeling. When you’re continuously learning from past security measures, you’re able to more effectively prevent incidents from occurring. 

Here are five statistical modeling tactics your federal agency needs to deploy to improve your predictive organizational security efforts: 

1. Use Only Clean, Formatted Data 

In order to leverage the full power of statistical modeling, you need to work with clean data. It should be in a format that’s easy for multiple users to work with and doesn’t require too many manual processing steps. Unformatted data that requires intensive parsing or formatting is difficult to keep clean, so aim for data that is easily entered or migrated into usable data fields. 

2. Make Your Models Reproducible 

Your statistical models should be backed by robust data science, which means any models you create (or hire others to create for you) should be easily reproducible. A different agency executive should be able to apply your same model to a different data set with the same variables and still produce valid results. Reproducibility isn’t just about academic integrity – it ensures your models don’t contain any errors with the potential to produce costly false positives. 

3. Conduct Exploratory Data Analysis 

Before you create your final statistical model, start with some exploratory data analysis. Compare high-level charts, find correlating variables and search for interesting patterns. If your federal agency is trying to model particular behaviors off of data indicators, take some time upfront to determine which variables and data points model those behaviors. This exploratory analysis helps you (or your professional services firm) create a more accurate and truly predictive model

4. Hire Data Science Experts 

While there are some aspects of statistical modeling you can perform yourself, the consequences of creating a poor model for your security operation could be drastic. Robust predictive modeling involves high-level data science and should be completed by professional advisors or contractors who specialize in data analysis for agency security. In addition, any potential contractor should hold a Top Secret Facility clearance and know how to handle sensitive data. 

5. Don’t Constrict Your Model To Past Events 

When you’re working to prevent rare security events (such as a security breach or insider threat incident), you might be tempted to build your model to prevent incidents that have happened similarly in the past. For example, if your predictive model is built to prevent the next Navy Yard shooting, it’s predictive power for other insider threat incidents could be wildly inaccurate. Your model must take past variables into account but remain flexible enough to predict yet-unknown future circumstances.  

Without tapping into the power of statistical modeling, your agency’s security efforts are at the mercy of unpredictable threats. However, when you leverage the power of predictive models, you proactively deter threats before they ever become full-scale incidents. 

Does your federal agency need to improve organizational security operations within a limited schedule and budget? Click below to download this e-book from Big Sky Associates and discover how to make process improvement efforts that are cost-effective for your budget. 

Download Your Free Report: The Ultimate Process Improvement Guide From Initial Data Analysis To Final Implementation Plan

Catch up with the rest of the elite data analysis series: