TODLER: A Transaction Ordering Dependency anaLyzER - for Ethereum Smart Contracts

Abstract

Smart contracts are programs with data (mutable state); stored on and executed by blockchain platforms. The transactions (or function invocations) dispatched to smart contracts often change their state. In the Ethereum blockchain, nodes (aka miners/validators) can schedule a set of transactions in any order in a block. Multiple transactions in a single block operating on a contract's shared state may yield different outcomes based on their execution order, thus creating a possibility for non-determinism and races between transactions. The resulting issue in Ethereum smart contracts is Transaction Ordering Dependency (TOD). Detecting a TOD requires identifying valid transactions affecting a contract's global/state variables which is equivalent to detecting read-after-write dependencies in race detection, and we expect it to be similarly nontrivial for human developers. In this paper, we identify various TODs, including a novel type previously undocumented in the literature. To detect these TODs, we propose an information flow analysis-based static analyzer, TODler. Our manual evaluation of 108 Ethereum smart contracts shows that TODler outperforms previously available approaches in terms of both run time and precision and also detects the novel TOD pattern identified in this paper.