CrossInspector: A Static Analysis Approach for Cross-Contract Vulnerability Detection

Abstract

With the development of blockchain technology, the detection of smart contract vulnerabilities is increasingly emphasized. However, when detecting vulnerabilities in inter-contract interactions (i.e., cross-contract vulnerabilities) using smart contract bytecode, existing tools often produce many false positives and false negatives due to insufficient recovery of semantic information and inadequate consideration of contract dependencies. We present CrossInspector, a novel framework for detecting cross-contract vulnerabilities at the bytecode level through static analysis. CrossInspector utilizes a trained Transformer model to recover semantic information and considers control flow, data flow, and dependencies related to smart contract state variables to construct a state dependency graph for fine-grained inter-procedural analysis. Additionally, CrossInspector incorporates a pruning method and two parallel optimization mechanisms to accelerate the vulnerability detection process. Experiments on our manually constructed dataset demonstrate that CrossInspector outperforms the state-of-the-art tools in both precision (97\%) and recall (96.75\%), while also significantly reducing the overall time from 16.34 seconds to 7.83 seconds, almost on par with the fastest tool that utilizes bytecode for detection. Additionally, we ran CrossInspector on a randomly selected set of 300 real-world smart contracts and identified 11 cross-contract vulnerabilities that were missed by prior tools.