一句话：提出交易级语义框架，检测智能合约日志伪造等漏洞、合成回滚攻击并解释链下不一致。
区块链交易常被链下系统通过调用轨迹、事件日志和存储修改来解释，但这些可观察产物可能因 revert 或误导性日志而与链上实际执行产生语义不一致。现有工具大多假设可观察效应与最终状态一致，忽视此类错配。本文提出面向智能合约安全分析的语义框架，建模并利用交易级语义，以识别真实执行效应与链下“意图无关”解释之间的不匹配，从而检测漏洞、合成攻击并解释链下不一致。作者计划实现三个工具：PEvent-Catcher 检测日志伪造漏洞，RollGain 合成基于回滚的状态恢复攻击，DeepTx 实时意图检测。
Blockchain transactions are often interpreted by off-chain systems through call traces, event logs, and storage modifications. However, these artifacts can diverge from the actual on-chain execution due to semantic mismatches caused by reverts or misleading logs. Existing tools largely assume consistency between observable effects and final state, overlooking semantic mismatches. We present a semantic framework for smart contract security analysis that models and leverages transaction-level semantics to detect vulnerabilities, synthesize attacks, and explain off-chain inconsistencies. Our approach identifies mismatches between real execution effects and intent-oblivious interpretations by off-chain systems. We plan to implement three tools: PEvent-Catcher for detecting log forgery vulnerabilities, RollGain for synthesizing rollback-based state-reverting attacks, and DeepTx for real-time intent detection.
DOI 原文 ·
@inproceedings{paperbot9551,
title = {Secure Transaction Semantics: Analysis, Vulnerability Detection, and Attack Modeling},
author = {Yixuan Liu},
booktitle = {ASE},
year = {2025},
doi = {10.1109/ase63991.2025.00398}
}