Server-Side ML Fraud Detection for Blockchain Wallets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Electronic payment systems face challenges in protecting users against fraud in virtual-currency-based transactions due to limited access to necessary data, making it difficult to detect and prevent fraudulent activities effectively.
Innovation Solution
An intelligent fraud detection system using machine-learning models is implemented to analyze transaction characteristics and provide real-time risk assessments, which can modify the user interface of a decentralized digital wallet to alert users of potential risks before approving transactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine-learning models are used to analyze transaction characteristics, then fraud detection accuracy is improved, but system complexity increases
Solution Approach 1:
The patent introduces a server as an intermediary component that hosts and executes the machine-learning models. This intermediary handles the complex computational tasks of analyzing transaction characteristics and generating risk scores, while the decentralized digital wallets can focus on their core functionality of facilitating transactions. The server acts as a specialized mediator that bridges the gap between simple transaction initiation and complex fraud analysis.
Solution Approach 2:
The patent extracts the complex fraud detection functionality from the decentralized digital wallet application and places it on a separate server infrastructure. By taking out the machine-learning model execution and training operations from the wallet application, the system reduces the complexity burden on end-user devices while maintaining sophisticated fraud detection capabilities on the server side.
2Loss of information
If real-time risk assessment is provided to users, then user awareness of fraud risks is improved, but transaction processing time increases
Solution Approach 1:
The patent implements preliminary action by pre-training machine-learning models on historical transaction data before deployment. The models are trained offline to recognize fraud patterns, so when actual transactions occur, the system can quickly apply these pre-learned patterns to generate risk scores without requiring time-consuming analysis during the transaction moment. This preliminary preparation enables fast real-time assessment.
Solution Approach 2:
The system applies partial action by providing risk assessments selectively rather than for every single transaction. The server can adjust the level of scrutiny based on transaction characteristics, applying more detailed analysis only when necessary indicators are present. This approach maintains user awareness of risks while avoiding unnecessary delays for low-risk transactions.
3Reliability
If the system analyzes multiple transaction characteristics, then fraud detection capability is improved, but data processing requirements increase
Solution Approach 1:
The patent segments the fraud detection process into distinct analytical components that evaluate different transaction characteristics separately. The machine-learning models are structured to process various features (transaction amount, frequency, counterparty information, device metrics) as separate input dimensions. This segmentation allows the system to analyze multiple characteristics systematically while managing data processing requirements through modular evaluation.
Solution Approach 2:
The system employs parameter changes by dynamically adjusting which transaction characteristics are analyzed based on the specific context. The machine-learning models can weight different parameters differently depending on the transaction type and risk indicators present. This flexible parameter approach enables comprehensive fraud detection capability while optimizing data processing requirements by focusing computational resources on the most relevant transaction attributes.
Data Source
AI summary
Disclosed herein are systems and methods for identifying fraud in blockchain-based transactions. In one method, a server detects an interaction by a decentralized digital wallet with a first electronic transaction protocol corresponding to a blockchain; executes a computer model (previously trained based on characteristics of electronic transaction protocols corresponding to the blockchain and electronic transactions corresponding to electronic transaction protocols) to determine a risk score; and in response to determining that the risk score for the requested electronic transaction is above a threshold level, causing at least one graphical element of a GUI associated with the decentralized digital wallet to provide an indication of a risk associated with the requested electronic transaction.