Data Management and AI for Blockchain Data Analysis:A Round Trip and Opportunities

Abstract

A blockchain platform is generally cohabited by human users, autonomous agents, cryptocurrencies, other digital assets, and decentralized protocols. As an example, consider the Ethereum ecosystem - currently the most actively used and the second-largest cryptocurrency network by market capitalization after Bitcoin. Ether is the native cryptocurrency of Ethereum that is transferred between accounts. Ethereum accounts are of two types: Externally owned accounts are controlled by users, whereas a contract account is controlled by a smart contract, which is an autonomous agent and can execute complex code across a decentralized network. For instance, smart contracts can define tokens that are digital assets in the blockchain platform. Decentralized applications (dApps) such as exchanges, wallets, and DeFi may combine multiple smart contracts and their protocols constitute a collection of rules that govern dApps in a decentralized blockchain platform. Complex interactions across various actors in blockchains generate massive-scale, dynamic, heterogeneous, and multi-modal data that are often publicly accessible and can be considered big data – an emerging trend since the past decade. Analysis of blockchain data using the latest data management and AI techniques is critical for the improvement of the blockchain technology, such as detecting and predicting trends, anomalies, e-crimes, and key actors. <br/><br/>In the first part of the talk, I shall discuss our recent work on blockchain data extraction and graph construction, graph mining, topological data analysis, and machine learning methods for various target applications such as detecting market manipulators in the blockchain world including the collapse of the stablecoin LunaTerra, Ethereum’s switch from Proof-of-Work (PoW) to Proof-of-Stake (PoS), and the stablecoin USDC’s temporary peg loss. In the second part, I shall showcase the contributions of blockchain technology in the growing ecosystem of data management and AI – in the form of diverse datasets, tools, novel challenges, and algorithms. I shall conclude by emphasizing future research directions such as cross-chain data analysis, combining signals from external sources, e.g., tweets and social media data about blockchains for holistic predictions, higher-order and multi-modal network analysis, designing of temporal machine learning and machine unlearning algorithms.