On-Chain Address Misclassification Detection Using Transaction Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The anonymous and low-information nature of blockchains does not provide sufficient assurances that a user has accurately classified themselves based on on-chain information, leading to potential misrepresentations and unsustainable risks in the long-term use of blockchain assets.
Innovation Solution
Implementing a system that uses machine learning models to analyze on-chain transactions and IP addresses to detect misclassifications by generating embedding vectors from transaction data, predicting use patterns, and updating records to flag misclassified addresses, thereby enforcing remedial actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning analysis is implemented to detect misclassifications, then detection accuracy is improved, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary analysis system that sits between the blockchain network and users. This intermediary uses machine learning models to analyze transaction patterns and IP addresses, acting as a mediator that provides detection capabilities without requiring changes to the underlying blockchain protocol or core system architecture.
Solution Approach 2:
The patent replaces manual verification mechanisms with automated machine learning analysis. Instead of relying on users to self-declare their identity or on simple cryptographic proofs, the system uses ML models to automatically analyze transaction behaviors, IP address patterns, and other data points to detect misclassifications.
2Reliability
If comprehensive transaction analysis is performed to detect misclassification, then detection capability is improved, but processing time increases
Solution Approach 1:
The patent implements a risk-based analysis approach where not all transactions are subjected to the same level of scrutiny. The machine learning model identifies suspicious patterns and focuses detailed analysis on transactions that exhibit misclassification characteristics, while allowing clearly legitimate transactions to pass through with minimal processing.
Solution Approach 2:
The system performs preliminary analysis of transaction data as it enters the network, using the machine learning model to identify potential misclassifications before they can cause harm. This allows for early detection and remediation, reducing the need for lengthy post-detection investigation processes.
Data Source
AI summary
A method and related system of operations that include retrieving a set of on-chain transactions of an on-chain address from a scan of on-chain transactions based on a request identifying the on-chain address and generating a sequence of embedding vectors by providing the set of on-chain transactions to a first machine learning model. The method further includes predicting a first use pattern of the on-chain address by providing the sequence of embedding vectors to a second machine learning model and determining whether the first use pattern matches with a second use pattern. The method further includes updating a record of the on-chain address in an off-chain database to indicate anomalous activity in response to a determination that the first use pattern does not match with the second use pattern.