Raze to the Ground: Query-Efficient Adversarial HTML Attacks on Machine-Learning Phishing Webpage Detectors

TL;DR

This paper introduces a novel, query-efficient black-box attack method that significantly degrades the performance of state-of-the-art ML-based phishing webpage detectors by using fine-grained, rendering-preserving HTML manipulations.

Contribution

It proposes a new set of HTML manipulations and an optimization algorithm to effectively bypass ML phishing detectors with minimal queries, improving attack effectiveness and robustness evaluation.

Findings

Attacks reduce detector performance with only 30 queries.

Fine-grained manipulations preserve webpage functionality and appearance.

Outperforms previous limited-effectiveness attacks.

Abstract

Machine-learning phishing webpage detectors (ML-PWD) have been shown to suffer from adversarial manipulations of the HTML code of the input webpage. Nevertheless, the attacks recently proposed have demonstrated limited effectiveness due to their lack of optimizing the usage of the adopted manipulations, and they focus solely on specific elements of the HTML code. In this work, we overcome these limitations by first designing a novel set of fine-grained manipulations which allow to modify the HTML code of the input phishing webpage without compromising its maliciousness and visual appearance, i.e., the manipulations are functionality- and rendering-preserving by design. We then select which manipulations should be applied to bypass the target detector by a query-efficient black-box optimization algorithm. Our experiments show that our attacks are able to raze to the ground the performance of current state-of-the-art ML-PWD using just 30 queries, thus overcoming the weaker attacks developed in previous work, and enabling a much fairer robustness evaluation of ML-PWD.