Research Projects

• Designing Adaptive Borrow-Lending Markets for Decentralized Finance

  • AFT 2024 (28%): Thinking Fast and Slow: Data-Driven Adaptive DeFi Borrow-Lending Protocol, (Mahsa Bastankhah, Viraj Nadkarni, Xuechao Wang, Chi Jin, Sanjeev Kulkarni, Pramod Viswanath)

  • (Under review) AgileRate: Bringing adaptivity and Robustness to Defi Lending Markets, (Mahsa Bastankhah, Viraj Nadkarni, Xuechao Wang, Pramod Viswanath)

Decentralized lending platforms operate through smart contracts on blockchains, where lenders deposit funds into liquidity pools and borrowers take out loans by overcollateralizing a second type of asset. Unlike traditional banks, which assess creditworthiness, DeFi lending is fully permissionless, meaning anyone can participate without needing a credit score. Because no trust is established between parties, over-collateralization is essential to ensure loan security. However, due to asset price volatility, determining the optimal over-collateralization factor requires careful consideration and algorithm design.

Another key distinction between DeFi markets and their traditional counterparts is their dynamic nature—while traditional markets adjust interest rates periodically, DeFi markets require adaptive mechanisms to respond to rapid fluctuations in supply and demand while staying resilient against adversarial manipulation. 

My research models supply and demand dynamics in decentralized lending markets and develops algorithms inspired by Kalman filtering methods to estimate user behavior model parameters from historical data and dynamically adjust interest rates and risk parameters, to optimize demand-supply balance while controlling insolvent debt risk.

While Thinking Fast and Slow emphasizes factorizing and quantifying various influences—such as risk and utility—that shape borrower and lender behavior, AgileRate takes a more practical approach, proposing a model of user behavior that is better supported by empirical data. Additionally, AgileRate provides a more extensive theoretical analysis of adversarial behavior, investigating to what extent malicious actors could exploit the system to manipulate interest rate to extreme values, potentially destabilizing the market.

• Algorithm Design for Payment Channel Networks

  • NDSS 2025 (20%): Alba: The Dawn of Scalable Bridges for Blockchains, Giulia Scaffino, Lukas Aumayr, (Mahsa Bastankhah, Zeta Avarikioti, Matteo Maffei)

  • FC 2023 (22%): R2: Boosting Liquidity in Payment Channel Networks with Online Admission Control, (Mahsa Bastankhah, Krishnendu Chatterjee, Mohammad Ali Maddah-Ali, Stefan Schmid, Jakub Svoboda, and Michelle Yeo)

  • Route discovery in private payment channel networks, (Z Avarikioti, M Bastankhah, MA Maddah-Ali, K Pietrzak, J Svoboda, M Yeo)

Payment Channel Networks (PCNs) are off-chain transaction networks built on blockchains. These networks significantly improve blockchain scalability by enabling transactions to be routed efficiently without requiring every payment to be recorded on-chain. Instead, only the opening and closing of channels are settled on the blockchain, ensuring users still benefit from the security, decentralization, and permissionlessness of blockchain systems.

Each of the above papers focuses on a different aspect of optimizing PCNs:

1. The first paper develops a secure protocol that enables the verification of off-chain transactions from a PCN on a separate blockchain. The protocol is rigorously analyzed under the Universal Composability (UC) framework and extensive-form game theory, ensuring robust security guarantees against byzantine adversaries.

2. The second paper designs an admission control algorithm for routers in PCNs, where nodes must decide which transactions to accept and route. Accepting transactions generates routing fees but also incurs unpredictable maintenance costs due to liquidity constraints. The paper introduces an online algorithm with a provable competitive ratio against an adaptive adversary, ensuring efficient transaction selection under uncertainty.

3. The third paper focuses on transaction routing in a privacy-preserving PCN, where nodes only have knowledge of their direct neighbors. The study proposes a routing mechanism that enables transactions to be efficiently relayed across the network while preserving user privacy, ensuring that intermediaries do not gain unnecessary visibility into the transaction path.

• Understanding Adversarial Manipulation in Consensus Algorithms

  • AFT 2023 (30%): F3B: a Low-Overhead blockchain architecture with per-Transaction Front-Running protection, (Haoqian Zhang, Louis-Henri Merino, Ziyan Qu, Mahsa Bastankhah, Vero Estrada-Galinanes, Bryan Ford)

  • FC 2023, International Workshops: Breaking Blockchain Rationality with Out-of-Band Collusion, (Haoqian Zhang, Mahsa Bastankhah, Louis-Henri Merino, Vero Estrada-Galiñanes, Bryan Ford)

Permissionless blockchains are vulnerable to adversarial manipulation due to their open nature. One major issue is Miner Extractable Value (MEV), where miners reorder, insert, or censor transactions for profit. The first paper proposes a protocol using threshold cryptography to encrypt transactions until finalization, preventing front-running and ensuring fair execution.

Another weakness is out-of-band collusion, where adversaries coordinate off-chain to manipulate consensus. Even with on-chain fairness mechanisms, external agreements can incentivize rational nodes to act maliciously. The second paper models this threat using game theory, showing how off-chain bribery enables double-spending and consensus manipulation, exposing vulnerabilities in blockchain security assumptions.