Streaming Classification of Distributed Ledger Transactions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for classifying distributed ledger-based activities, such as cryptocurrency transactions, face challenges in detecting illicit transactions in real-time due to the dynamic nature of cryptocurrency networks and the anonymous nature of some ledgers, leading to inefficiencies in fraud detection and classification.
Innovation Solution
The implementation of a method that combines Graph Convolutional Networks (GCN) or Random Forest classifiers with exponential time sampling to classify transactions in real-time, allowing the system to adapt to changes in the network and improve detection of illicit activities by sampling labels and features effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning algorithms (Random Forest, XGBoost, GCN) are used to classify transactions, then detection accuracy improves, but real-time processing capability deteriorates
Solution Approach 1:
The system performs preliminary actions by pre-training the GCN model offline with historical transaction data to learn temporal patterns and network structures. This pre-computed knowledge is then applied in real-time inference without requiring complex computations during transaction processing, thus achieving both high accuracy and real-time performance.
Solution Approach 2:
The patent segments the classification task into distinct phases: offline model training phase where GCN learns from historical data, and online inference phase where the trained model processes transactions in real-time. This segmentation allows computationally intensive operations to be performed offline while maintaining fast real-time processing capability.
2Measurement precision
If comprehensive transaction data is collected for training, then model accuracy improves, but data processing time and complexity increase
Solution Approach 1:
The system performs preliminary data processing and feature extraction offline before model training. Transaction data is pre-processed to extract relevant features and construct graph representations in advance, reducing the computational burden during real-time processing and minimizing data processing time while maintaining model accuracy.
Solution Approach 2:
The patent extracts and focuses on the most relevant features from comprehensive transaction data using the GCN model's attention mechanisms and graph convolution operations. By extracting only the essential features needed for classification, the system reduces data processing complexity while maintaining high detection accuracy.
3Reliability
If the system adapts to temporal changes in cryptocurrency networks, then detection reliability improves, but system complexity increases
Solution Approach 1:
The patent implements dynamic adaptation by using GCN to learn temporal evolving patterns in cryptocurrency transaction networks. The model dynamically adjusts its parameters and representations based on changing network structures and transaction patterns over time, allowing the system to adapt to temporal changes while maintaining a relatively simple architecture through the inherent capabilities of graph neural networks.
Data Source
AI summary
Systems and methods for streaming classification of distributed ledger-based activities are disclosed. In one embodiment, a method for streaming classification of distributed ledger-based activities may include: receiving, at a detection computer program executed by a server and from a distributed ledger network, information for a plurality of transactions; receiving, from a database and at the detection computer program, node information that identifies a plurality of nodes as involved in illicit transactions, as being involved in licit transactions, or unknown; applying, by the detection computer program, exponential time sampling to sample labels and associated features; training, by the detection computer program, a classifier; receiving, by the detection computer program, a streaming transaction from the distributed ledger network; predicting, by the detection computer program, a classification for the streaming transaction using the trained classifier; and outputting, by the detection computer program, an alert based on the prediction.