Decentralized Finance (DeFi) introduces a new financial landscape where user anonymity and diverse token interactions pose unique analytical challenges. In this talk, I explore the Ethereum token transfer network, which, unlike Bitcoin, hosts complex inter-token relations such as ERC-20 and NFT exchanges. By leveraging network science, I address gaps in existing transaction analysis methods that either depend on sparsely labeled data or overlook distinct DeFi user behaviors.

My approach employs graph motif-based fingerprints that can effectively infer transaction functions, enabling a clearer differentiation of user groups and token dynamics. I present a case study of Alameda Research’s activity leading up to the FTX collapse, where analysis of multi-token flows and node centralities uncovers shifts in token distributions and user centrality. These findings offer critical insights into the influence of high-profile accounts and evolving transaction dynamics within DeFi.

Beyond transaction networks, DeFi lending protocols like Compound and AAVE face risks analogous to traditional financial contagions. Collaborating with economists, I utilize blockchain’s transparency to model systemic risks, simulating price shocks and cascading liquidations within DeFi lending pools. Through these simulations, we identify risk-prone lending pools and behaviors that may destabilize protocols. My ongoing work integrates user-level data in a bipartite network to detect contagion-prone users and lending pools.

This talk will emphasize how network science and real-time data applications in blockchain offer a powerful lens for analyzing the dynamics and systemic risks inherent to DeFi ecosystems.