Risk Assessment for Loan Defaults in Decentralized Finance(DeFi) Lending Platforms

Abstract

MakerDAO is a decentralized lending protocol providing crypto-backed loans with no intermediaries, backed by volatile assets such as Ethereum (ETH). Loans are liquidated if the value of the collateral dips below a threshold. DeFi compared to traditional finance does not have standardized risk models and default is difficult to model. Earlier models such as Poisson Process and Brownian Motion have the unrealistic premise of constant volatility, which makes them less useful in rapidly fluctuating crypto markets. This paper introduces a Geometric Brownian Motion (GBM) model with rolling volatility to capture real-time market dynamics. The model learns to adapt to prevailing price trends by estimating volatility with a rolling window and enhances the accuracy of default risk estimation. Results indicate that rolling volatility increases the predictive ability of GBM, providing a robust solution to credit risk management in DeFi platforms. The GBM with rolling volatility has 0.006 root mean square error and 0.008 mean absolute error.