James Hsin-yu Chiang

Sarah Allen, Pranay Anchuri, James Austgen et al.

The intersection of crypto x AI is spawning papers, products, online posts, and companies. All the surrounding buzz, though, obscures what exactly has been done, what the opportunities and challenges are, and what open questions deserve attention. This survey paper asks what AI can do for blockchain-based technologies (broadly construed as "crypto") (crypto x AI), and vice versa (AI x crypto). We systematize existing work, summarize key takeaways, highlight open research questions, and offer a perspective on pervasive industry misconceptions, concluding that AI and crypto are still in the very early stages of meaningful integration.

Massimo Bartoletti, James Hsin-yu Chiang, Alberto Lluch-Lafuente

Automated Market Makers (AMMs) are decentralized applications that allow users to exchange crypto-tokens without the need for a matching exchange order. AMMs are one of the most successful DeFi use cases: indeed, major AMM platforms process a daily volume of transactions worth USD billions. Despite their popularity, AMMs are well-known to suffer from transaction-ordering issues: adversaries can influence the ordering of user transactions, and possibly front-run them with their own, to extract value from AMMs, to the detriment of users. We devise an effective procedure to construct a strategy through which an adversary can maximize the value extracted from user transactions.

Massimo Bartoletti, James Hsin-yu Chiang, Alberto Lluch-Lafuente

Lending pools are decentralized applications which allow mutually untrusted users to lend and borrow crypto-assets. These applications feature complex, highly parametric incentive mechanisms to equilibrate the loan market. This complexity makes the behaviour of lending pools difficult to understand and to predict: indeed, ineffective incentives and attacks could potentially lead to emergent unwanted behaviours. Reasoning about lending pools is made even harder by the lack of executable models of their behaviour: to precisely understand how users interact with lending pools, eventually one has to inspect their implementations, where the incentive mechanisms are intertwined with low-level implementation details. Further, the variety of existing implementations makes it difficult to distill the common aspects of lending pools. We systematize the existing knowledge about lending pools, leveraging a new formal model of interactions with users, which reflects the archetypal features of mainstream implementations. This enables us to prove some general properties of lending pools, such as the correct handling of funds, and to precisely describe vulnerabilities and attacks. We also discuss the role of lending pools in the broader context of decentralized finance.

James Chiang

Smart contracting protocols promise to regulate the transfer of cryptocurrency amongst participants in a trustless manner. A safe smart contract implementation should ensure that each participant can always append a contract transaction to the blockchain in order move the contract towards secure completion. To this goal, we propose Bitcoin Trace-Net, a contract verification framework which generates an executable symbolic model from the underlying contract implementation. A Trace-Net model consists of a Petri Net formalism enriched with a Dolev-Yao-like actor knowledge model. The explicit symbolic actor knowledge model supports the verification of contracts featuring cryptographic sub-protocols, which may not be observable on the blockchain. Trace-Net is sufficiently expressive to accurately model blockchain semantics such as the delay between a transaction broadcast and its subsequent confirmation, as well as adversarial blockchain reorganizations of finite depths, both of which can break smart contract safety. As an implementation level framework, Trace-Net can be instantiated at run-time to monitor and verify smart contract protocol executions.