Socure Solution Insights: Preventing Digital Fraud Attacks with Enhanced Email RiskScore Models and Generative AI

Battling Large-Scale Digital Fraud Attacks in Finance and Retail: A Case for Socure’s Sigma Identity Fraud Solution

How to Stay Ahead of Digital Payments Fraud with Socure Account Intelligence

What Every Fraud Executive Should Be Aware of in Times of Banking Turmoil

As digital platforms become more ubiquitous, financial institutions and merchants face growing threats from large-scale fraud attacks. To combat this digital fraud epidemic, many organizations rely on <a href="https://www.socure.com/products/sigma-identity-fraud">Socure’s Sigma Identity Fraud solution</a> paired with <a href="https://www.socure.com/products/email-risk">Email</a>, <a href="https://www.socure.com/products/address-risk">Address</a> and <a href="https://www.socure.com/products/phone-risk">Phone RiskScores</a>.
Now, we’re taking fraud prevention to the next level with Email RiskScore v10.1.
Powered by generative AI technology, v10.1 reduces false positives by ~4% and increases name-email correlation by 2-6% over earlier versions. It brings enhanced email risk scoring through:
<ul>
<li>Detecting nonsensical and fake emails using natural language processing to measure email handle “perplexity”. Gibberish handles with no contextual meaning are flagged.</li>
<li>Analyzing billions of real email handles to understand legitimate construction patterns. Handles diverging from these expected patterns raise risk levels.</li>
<li>New techniques identifying fake/test emails, risky domains, and fuzzy matching to correlate emails with identities.</li>
</ul>
The upgrades provide greater accuracy in pinpointing high-risk emails associated with fraud attacks. With explainable risk scores and reason codes, customers also gain more transparency into decisions.
<h2>Fighting Back with Generative AI</h2>
At the core of v10.1 is Socure’s use of generative AI. The team trained a neural language model on billions of email handles from real people, learning the typical patterns and structures people follow when creating addresses.
The model then assigns a “perplexity” score, indicating how improbable an email handle’s word or character sequence is compared to legitimate patterns. High perplexity suggests a fraudulent, randomized handle. The sequence information improves Socure’s ability to mimic human judgment in assessing email handle legitimacy.
Combined with robust analytics on email domains, names, and other identity attributes, it produces Socure’s most advanced email risk scoring to date. Ongoing generative AI innovation will only make these models stronger over time.
With Email Risk Score v10.1, financial institutions and merchants have a powerful new tool to fight the growing threat of digital fraud. By enhancing risk scoring through the latest AI, Socure enables customers to stop more fraud, reduce false positives, and improve approval rates.

Socure Solution Insights: Preventing Digital Fraud Attacks with Enhanced Email RiskScore Models and Generative AI

Emily Saitta

Related Posts