Speaker: Dr. Cuneyt Akcora
From: UCF College of Business, AI for Finance
Abstract
Blockchains allow pseudo-anonymous transactions, which has made it easier to create a payment ecosystem used worldwide. However, the ease of blockchain use has also attracted e-crime actors with malicious activities ranging from sextortion to ransomware demands and darknet market sales. E-crime has already resulted in significant economic losses and societal harm across different sectors, ranging from local governments to health care.
Most e-crime operators use Bitcoin and a few cryptocurrencies. Although Bitcoin transactions are permanently recorded and publicly available, current approaches for detecting e-crime depend only on a couple of heuristics or tedious information-gathering steps (e.g., manually collecting ransomware-related Bitcoin addresses). Furthermore, adversaries (e-crime operators) change their behavior in time to avoid detection by Machine Learning (ML) models that are used by blockchain data analytics companies and law enforcement agencies.
We propose a novel approach, the Chainlet methodology, as an effective data analytics framework for conducting graph ML tasks on blockchains. Our approach uses graph substructures to construct transaction networks and creates topological embeddings of both blockchain addresses and temporal blockchain graphs. We use Chainlets to create interpretable results for the adversarial settings, enabling us to better combat e-crime.
For more info, please follow this link.
Read More