The Simility Blog
Quickly Adapting to New Fraud Attacks by Easily Updating Rules and Machine-Learning Models
Kedar SamantMay 08, 2018
As you know, organizations are seeing constant new fraud attacks on their systems. New account fraud increased by 70% over the course of 2017 and losses from account takeover fraud (ATO) grew by 120% between 2016-2017, according to Javelin Strategy & Research 2018 Identity Fraud Study. Ouch!
To keep up with these new attacks, fraud analysts and data scientists need to update their rules and machine learning (ML) models. Such changes enable systems to detect fraud attempts across the organization.
But making these changes isn’t always easy. Any live system usually has hundreds of rules in place, making it difficult to know which rules are effective, which might be superfluous, which have been changed, when those changes occurred, and who enacted them. Managers might be able to make the distinction. But depending on staffing, it could take lots of time, money, and resources to do so.
What Organizations Need
When it comes to adding new rules and ML models, organizations need governance that includes testability functions for what happens if a fraud analyst changes a rule strategy. Such insight is crucial in that it allows organizations to manage the effectiveness of a change and foresee how a change will affect their systems’ fraud detection rates. The hope is that a new rule can help organizations detect more attacks without generating false positives and/or conflicting with other effective rules/models.
Organizations also need to have all of this testing, auditing, change measure, and other governance structures available in a simple UI. That way, line managers, vice presidents (VPs) of fraud, risk managers, and possibly even C-level executives can understand how each rule and ML model helps protect the organization.
A Step Ahead with Simility
Simility understands organizations’ desire to easily add and adjust effective rules. Its platform uses a simple UI allowing for full governance testing, rule optimization, and data management as a service. In essence, the solution automatically provides rule-change management by logging and generating alerts or audit reports for each change to the rule set, thereby saving time and resources as well as improving data security.
Administrators can also use Simility to analyze changes to rules and machine learning models by looking back at previous data sets. This information helps illustrate the impact of each change, generating better results that help organizations improve their business outcomes and please stakeholders. Simility also helps analysts by automatically tuning and optimizing strategies and presenting them to analysts with explainability and decision clarity. With this knowledge, analysts can run champion/challenger rule sets to ease decision-making about changes. If errors arise, organizations can use Simility to quickly and efficiently roll back changes.
Interested in learning more about how Simility can help your organization add and adjust effective rules and ML models? Schedule a meeting and demo.