Simple Perturbations Subvert Ethereum Phishing Transactions Detection: An Empirical Analysis

TL;DR

This study demonstrates that simple perturbations can significantly undermine Ethereum fraud detection models, revealing vulnerabilities and suggesting mitigation strategies to improve robustness.

Contribution

The paper provides an empirical analysis of the susceptibility of common machine learning models to simple adversarial attacks in Ethereum transaction detection.

Findings

Simple attacks drastically reduce model accuracy

Different algorithms show varying vulnerability levels

Mitigation strategies can improve robustness

Abstract

This paper explores the vulnerability of machine learning models, specifically Random Forest, Decision Tree, and K-Nearest Neighbors, to very simple single-feature adversarial attacks in the context of Ethereum fraudulent transaction detection. Through comprehensive experimentation, we investigate the impact of various adversarial attack strategies on model performance metrics, such as accuracy, precision, recall, and F1-score. Our findings, highlighting how prone those techniques are to simple attacks, are alarming, and the inconsistency in the attacks' effect on different algorithms promises ways for attack mitigation. We examine the effectiveness of different mitigation strategies, including adversarial training and enhanced feature selection, in enhancing model robustness.