Deloitte's Crypto Currency Transaction Tracking Tool
Crypto means big business – and a big deal for compliance
Using public keys to track transactions is becoming challenging: Whereas criminals previously posted public keys on social media, it can be observed that they are much more cautious about this practice now. With this type of information (OSINT) drying up, it is increasingly problematic that the majority of tracking tools on the market rely on it as their source while the available database is getting steadily worse.
In response to this problem, Deloitte Consulting
Germany and Deloitte Forensic Germany have chosen a new approach for the
development of their own tracking tool.
The two teams have set up a Neo4J graph database,
which in addition to ledger information, i.e. the actual transaction data,
contains publicly available cluster (“which transactions belong to which
wallet”) and labelling (“who owns the wallet”) information.
The tracking tool adds in further information, such as
the “unspent amount” of a transaction, that feeds into algorithms and
heuristics, which are probability calculations for transaction patterns, to
recognize affiliations and identify anomalies.
In a comprehensive analysis, the monitoring tool also processes data such as the sums transferred, network times, transaction activities, and chronological patterns, and maps these to the usage types of different actors such as stock exchanges, investors, and end users, as well as technical tools such as mixers.
The central core of the analysis is based on so-called “supervised learning”. Learning algorithms make hypotheses, i.e. they map input values to assumed output values. The goal is to predict the classification (“to solve the classification problem”) of the transaction wallet and map the wallet identity as accurately as possible. In iterative training and test cycles, the associations between transaction and wallet (clustering) and between wallet and identity (labelling) improve over time.
Methods of logistic regression, conditional probability application, likelihood functions, perceptron algorithms are then applied. Extensions such as (naïve) Bayesian classifiers, next-neighbor classification, discrimination analysis, artificial neural networks, and the extension of random walks or random movements are planned for the next development/expansion stage of the tool. These will make conspicuous patterns and irregularities in transactions visible even when information is sparse.
The Deloitte tracking tool can be used instantly without installation, as it is purely web-based. Furthermore, it can be integrated easily into existing transaction monitoring and other peripheral systems due to its universal API interface.
Have we caught your interest? Please reach out to us, we are more than happy to showcase the tool to you and discuss potential integration solutions.