Skip to main content
We Need AI in Cybersecurity

Last month I spoke on a panel at Money20/20 in Las Vegas entitled, “Using Artificial Intelligence & Data Analytics for Managing Fraud Risk & Data Security.”  Despite being the second-to-last time slot, it was well attended by people who are working to increase their use of machine learning in their fraud and risk applications.

For me, this provided an opportunity to reflect on the more than two decades that card issuers have been using Falcon Fraud Manager to make payment authorization decisions through a highly predictive neural network score.  It is immensely satisfying to me that the machine learning and more than 100 patents in fraud analytics that FICO developed have become the benchmark for detecting payment card fraud, protecting 2.5 billion cards worldwide.

Where the payment card fraud industry was in the early 1990s (before Falcon) is very similar to where we are today in the cybersecurity realm. There’s an over-reliance on heuristics and signature methods, which are ineffective.  Armies of cyber analysts (which don’t exist) and mountains of rules won’t be enough in the near-term and certainly won’t work in the long-term to prevent loss of sensitive data. Add the Internet of Things and Bring Your Own Device to the mix, and we’re looking at a future of breaches and data theft that continues to fuel the fraud machine with stolen PII/card details.

To address the out-of-control cybersecurity problem, we need the widespread adoption of predictive machine learning analytics. Today, your card is better protected than your organization.

The cybersecurity analytics FICO is developing are based on self-learning technologies such as:

  • Self-calibrating analytics
  • Adaptive analytics based on case review data
  • Self-classification of users and assets through adaptive soft clustering techniques, including collaborative profiling
On the Money20/20 panel, we talked about how these technologies represent the evolution in fraud analytics over the last two decades: away from strictly supervised models and toward those that leverage historical data and self-learning techniques.  The incorporation of these analytic techniques in FICO Falcon Fraud Manager 6 provides us with demonstrable results of the effectiveness of these techniques that are core elements of successful cybersecurity analytics.

The future is bright for tackling cybersecurity with highly predictive scores. What worked in fraud will work in cyber — and we need to start now.

For more information, watch this short video I recorded with Business Reporter:

related posts