Dynamic Risk Parameter Adjustment for Online Payment Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing risk identification methods for online transactions are plagued by high costs and significant delays due to the need for frequent manual updates and offline retraining of risk identification systems, which are inadequate in addressing emerging unauthorized access methods.
Innovation Solution
A method and device for dynamically adjusting risk parameters in real-time based on operation data from current transactions, converting transaction data into operation hopping sequences, and updating risk parameters corresponding to each operation hopping event, allowing for timely and online updates without manual intervention or offline retraining.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If risk identification rules are manually maintained and updated, then the system can adapt to new risks, but the cost and time required for updates become excessively high
Solution Approach 1:
The risk identification system automatically updates its own rules by learning from transaction data without requiring manual intervention. The system performs self-training using machine learning algorithms on historical and real-time transaction data, enabling it to adapt to new risks autonomously and continuously
Solution Approach 2:
The system implements a feedback mechanism where transaction outcomes (safe/unsafe classifications) are fed back into the risk identification system. This feedback loop allows the system to continuously learn from actual transaction results and adjust its risk parameters accordingly, improving adaptability over time
2Measurement precision
If the risk identification system is retrained offline, then the system can learn from historical data, but the process takes months to complete causing significant delays
Solution Approach 1:
The system transitions from static offline retraining to dynamic online learning. Risk parameters are adjusted in real-time as transactions are processed, allowing the system to maintain high accuracy while adapting continuously without interruption to transaction processing
Solution Approach 2:
The system performs preliminary risk parameter adjustments based on real-time transaction patterns before formal offline retraining is needed. This allows the system to proactively adapt to emerging risks while maintaining the option for comprehensive periodic retraining
3Extent of automation
If artificial intelligence and machine learning are used for risk identification, then the system can automatically identify risks, but the system becomes dependent on known risks and cannot handle unknown threats
Solution Approach 1:
The risk identification system segments risk parameters into multiple independent dimensions (e.g., transaction amount, time, location, device characteristics). This segmentation allows the system to analyze and adjust individual risk factors independently, enabling better detection of novel threat patterns that don't fit existing known risk profiles
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for payment services are provided. One of the methods includes: receiving a payment request, the payment request containing a user identifier and operation data generated in a current transaction corresponding to the user identifier; determining, according to the operation data, at least one operation hopping sequence in the current transaction, the operation hopping sequence containing at least one operation hopping event; obtaining, from a set of risk parameters corresponding to the user identifier, a risk parameter corresponding to the operation hopping event in the operation hopping sequence, risk parameters in the set of risk parameters being obtained by adjusting risk parameters used for risk identification in a previous transaction according to operation data generated in the previous transaction corresponding to the user identifier; and adjusting the obtained risk parameter according to the at least one operation hopping sequence.