Email Pattern Detection for Real-Time New Account Fraud
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to effectively detect and prevent new account fraud (NAF) in distributed computer systems, particularly in content delivery networks (CDNs), as fraudsters exploit email address patterns to create multiple accounts, obscuring their identity and maximizing personal gain.
Innovation Solution
A machine learning model with two computational branches is trained on email structures, processing the prefix and suffix portions of email addresses using a multi-head self-attention mechanism and convolutional neural network, respectively, to generate a score indicating potential fraud, enabling real-time detection and prevention of fraudulent account creation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional email-based detection methods are used, then implementation is simple, but detection accuracy is low due to fraudsters exploiting email patterns
Solution Approach 1:
The email address is segmented into two distinct parts: the prefix (username) and the suffix (domain). The patent applies different computational branches to each part - a self-attention mechanism for the prefix to capture character-level patterns and relationships, and a convolutional neural network for the suffix to detect domain-based patterns. This segmentation allows the system to analyze each component with specialized techniques, improving overall detection accuracy while maintaining manageable complexity through modular architecture.
2Reliability
If real-time detection is implemented, then fraud prevention is effective, but processing latency increases
Solution Approach 1:
The system performs preliminary action by pre-training the machine learning model on extensive email pattern data before deployment. During runtime, the pre-trained model can quickly evaluate new email addresses without requiring complex real-time analysis. The model has already learned to identify fraudulent patterns during the offline training phase, enabling fast real-time predictions that maintain low latency while ensuring reliable fraud detection.
3Ease of operation
If multiple email variations are allowed during registration, then user convenience is improved, but fraudsters can create multiple accounts using the same email
Solution Approach 1:
The system implements feedback by analyzing the relationship between different email prefix variations and their corresponding suffixes. The self-attention mechanism in the computational model detects patterns where fraudsters use multiple prefixes (e.g., different capitalizations, added characters) with the same suffix to create multiple accounts. The model learns from training data to identify these coordinated variations and flags them as potentially fraudulent, maintaining registration convenience for legitimate users while preventing abuse.
Data Source
AI summary
A method, apparatus and computer program product for real-time new account fraud detection and prevention. The technique leverages machine learning. In this approach, first and second computational branches of a machine learning model are trained jointly on a corpus of emails. Following training, an arbitrary email is received. The arbitrary email is then applied through the computational branches of the machine learning model. The first branch has an attention layer, and the second branch has a convolutional layer. The outputs of the branches are aggregated into an output that is then applied through another self-attention layer to generate a score. Based on the score, the arbitrary email is characterized. If the email is characterized as fraudulent, a mitigation action is taken.