GasTrace: Detecting Sandwich Attack Malicious Accounts in Ethereum

这段摘要讲述以太坊交易数据的公开透明性容易被恶意攻击利用，其中夹层攻击是一种通过前/后置交易操纵市场价格从而获利的攻击方式。为了识别和预防夹层攻击，作者提出了一个级联分类框架GasTrace。GasTrace通过分析各种交易特征来检测恶意账户，特别是通过燃气特征的分析和建模。在初始分类中，作者利用支持向量机（SVM）和径向基函数（RBF）核来生成账户的预测概率，并进一步构建详细的交易网络。随后，行为特征通过图注意力网络（GAT）技术在第二次分类中捕获。通过级联分类，GasTrace可以分析和分类夹层攻击。作者的实验结果表明，GasTrace具有显着的检测和生成能力，对于识别夹层攻击账户，准确率为96.73％，F1得分为95.71％。

GasTrace: A Cascade Classification Framework to Detect Sandwich Attacks in Ethereum

GasTrace uses a cascade classification framework to analyze various transaction features and detect malicious accounts that execute sandwich attacks in Ethereum. It utilizes SVM and GAT techniques to capture gas and behavior features and achieve a remarkable detection and generation capability.

GasTrace achieves an accuracy of 96.73% and an F1 score of 95.71% for identifying sandwich attack accounts. The framework is designed to analyze and model gas and behavior features, construct a detailed transaction network, and classify sandwich attacks. The authors also provide an open-source implementation of GasTrace.

Related studies include 'Sandwich Detection: A Machine Learning Approach to Detect Sandwich Attacks in Ethereum' by L. Li et al. and 'Detecting Sandwich Attacks in Ethereum with Graph Convolutional Networks' by Y. Zhang et al.