Blockchain Fraud Prediction via Machine Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for identifying fraudulent transactions in blockchain-based cryptocurrencies are cumbersome and computationally expensive due to the large number of inputs and outputs in each transaction, making real-time fraud detection challenging.

Innovation Solution

A computer-implemented method using machine learning to predict fraudulent transactions by training a model with transaction data from previous transactions, including information about inputs and outputs, to improve prediction accuracy and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If heuristic methods are used to label transactions as fraudulent based on addressee characteristics, then fraud detection capability is achieved, but the process becomes cumbersome and untenable in real-time

Engineering Contradiction:
Improvefraud detection capabilityVSAvoidreal-time processing speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent replaces manual heuristic analysis with an automated machine learning model that processes transaction data. The model takes transaction features (input addresses, output addresses, amounts, timestamps) and automatically predicts fraud likelihood, eliminating the need for manual heuristic evaluation while maintaining detection capability.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent creates a trained machine learning model that captures fraud patterns from historical data. This model serves as a copied representation of fraud detection expertise, enabling rapid prediction on new transactions without requiring real-time human analysis or complex heuristic rule evaluation.

Inventive Principle:
Principle #26Copying

2Measurement precision

If analysis is performed on all inputs and outputs of a transaction (up to 2000 each), then comprehensive fraud detection is achieved, but computational cost becomes prohibitive

Engineering Contradiction:
Improvefraud detection accuracyVSAvoidcomputational cost
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent extracts only the most relevant features from transaction data for model input. Instead of analyzing all 2000 inputs and outputs in detail, the model selects key features such as transaction amount, number of inputs/outputs, address characteristics, and temporal patterns, significantly reducing computational requirements while maintaining detection accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by focusing analysis on the most suspicious or relevant aspects of transactions. The machine learning model is trained to identify fraud patterns without requiring exhaustive analysis of every transaction component, achieving effective detection with reduced computational effort by concentrating on high-risk indicators.

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If machine learning models are trained with extensive transaction data including previous transactions, then prediction accuracy improves, but training and processing time increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidtraining and processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary action by training the machine learning model in advance using historical transaction data. The model learns fraud patterns from previous transactions during an offline training phase, so that when new transactions need to be evaluated, the model is already prepared and can make rapid predictions without requiring real-time analysis of historical data.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4481655A1Predicting whether a transaction of a digital currency stored in a blockchain is fraudulent
Publication Date: 2024.12.25 MASTERCARD INT INC
  • EP4481655A1 patent drawingFigure 1
  • EP4481655A1 patent drawingFigure 2
  • EP4481655A1 patent drawingFigure 3

AI summary

A computer implemented method of training a model, using a machine learning process, to predict whether a transaction of a digital currency stored in a blockchain is fraudulent, comprises: obtaining (202) transaction data for a first transaction of first funds in the digital currency, wherein the transaction data further comprises information related to a second transaction of the first funds that preceded the first transaction. The method further comprises labelling (204) the transaction data for the first transaction according to whether the first transaction was fraudulent and using (206) the transaction data and the label as training data with which to train the model.