Center for Digital Finance and Technology Announces New Research Grants
The Center for Digital Finance and Technology (CDFT) at Columbia Engineering is pleased to announce a third year of CDFT Research Grants. Established in 2022, the Center is focused on advancing the digital transformation of financial services for higher efficiency and security, increased accessibility, and greater social responsibility.
The projects selected this year tackle state-of the art methodologies for automated market makers as well as new proposed technologies in decentralized systems for privacy, security, and performance.
Awarded proposals are:
Stochastic Routing for Automated Market Makers
PI: Ciamac Moallemi, William von Mueffling Professor of Business
In a world of many AMMs and complex hooks, how should one route a trade? This proposal seeks to develop a new methodology for routing trades between AMMs, based on ideas from online learning, reinforcement learning, and meta-learning. The key insight we develop is that the optimal routing decision should equate the future expected marginal cost across AMMs at the time when the trade is executed, which incur some slippage relative to the costs at planning time. We seek to develop methods for estimating this slippage quantity, and to use it to develop better routing algorithms. We also seek to develop conditional routing algorithms that will allow for some degree of re-optimizing the routing decision at execution time. Finally, we discuss empirical approaches for estimating the optimality of existing routing strategies. Our preliminary empirical results suggest that existing routing strategies such as that used by the Uniswap Universal Router are incurring significant costs relative to the optimal strategy.
Variable Fee Designs for Automated Market Makers
PI: Marcel Nutz, Professor of Statistics
Passive liquidity providers (LPs) offering liquidity in automated market makers (AMMs) are exposed to
predictable losses due to adverse selection. Reality has shown that static trading fees often do not sufficiently compensate LPs for such losses. Variable fees, that change automatically with market conditions, have been proposed as a possible improvement. Our goal is to provide guidance on how variable fees should be structured and what consequences can be expected if they are implemented.
To that end, we build a simulation environment for an ecosystem where noise traders and arbitrageurs interact on a centralized exchange (CEX) and an AMM. Prices on the CEX evolve exogenously and noise trader flow is routed to the venue offering the best price inclusive of fees. Arbitrageurs act whenever roundtrip trades are profitable. LPs provide liquidity at the initial time and are otherwise passive (but may hedge their exposure). Simulating over a time interval, we compute key metrics of interest, in particular the profit or loss of a representative LP, as a function of numerous characteristics including the fee on the AMM, price volatility on the CEX, noise trader demand, and others. Analyzing this functional relationship, we infer which characteristics are the primary determinants of LP profitability (and of other desirable features). By optimizing the fee as a function of the other characteristics, we find the theoretical variable fee mechanism that optimally compensates LPs. Based on these insights, we will then propose and test concrete, implementable functional forms of variable fees that appropriately compensate LPs while also balancing other practical requirements. To the best of our knowledge, this will be the first work to systematically study variable fee mechanisms in a quantitative and data driven way.
Order flow auction under the Proof of Stake protocol – design and economics
PI: Wenpin Tang, Assistant Professor of Industrial Engineering and Operations Research
Financial industry has revolutionized in the last decade, and witnessed the emergence of financial
technology (FinTech) that transforms our everyday lives. The current project, aims to take further steps in the blockchain innovation – – the greatest invention in the FinTech era. The blockchain, and its crypto-carriers as the Bitcoin and Ethereum are widely believed to be the next-generation digital transaction platforms, and enable decentralizing business models. A unique feature of the blockchain is decentralization, which relies on consensus mechanisms. There are two widely used consensus mechanisms: Proof-of-Work (PoW) and Proof-of-Stake (PoS). PoW requires substantial computational power to solve cryptographic puzzles, making it highly energy-intensive. In contrast, PoS is more energy-efficient as it selects validators based on their stake rather than computational effort. The goal of this project is to provide further understanding of decentralization backing the PoS protocol, with novel design components as order flow auction (OFA) and Proposer-builder separation (PBS). We propose a novel game-theoretic approach, which concurs with the decentralized nature of the PoS blockchain. The study requires development of various tools from stochastic processes, convex optimization, and discrete mathematics, which are of independent interest.
ZKFUZZ: Practical Fuzzing for Zero-Knowledge Circuits
PI: Junfeng Yang, Professor of Computer Science
Zero-knowledge (ZK) proofs are a transformative technology in decentralized systems for privacy, security, and performance. Yet existing ZK circuits are rife with vulnerabilities that have led to billions of dollars loss in the Blockchain world. We propose to build ZKFUZZ, a novel dynamic fuzzing tool designed to detect and mitigate critical vulnerabilities in zero-knowledge circuits. By jointly fuzzing both inputs and program code, ZKFUZZ systematically exposes both under-constrained (invalid program traces are accepted) and over-constrained (valid program traces are rejected) flaws—issues that current static analysis and formal verification methods struggle to address effectively. Our approach combines rigorous theoretical formulation with practical, automated testing algorithms to enhance the security and reliability of zero-knowledge circuits in decentralized systems, particularly within the Ethereum ecosystem. Our prototype has already found 38 previously unknown bugs, 18 of which have been confirmed by developers including 11 in zk-regex, a popular regex verification circuit library, earning bug bounty.