Explanation-Guided Backdoor Poisoning Attacks Against Malware Classifiers

TL;DR

This paper explores how explainable AI techniques can be used to craft effective backdoor poisoning attacks on malware classifiers, highlighting vulnerabilities in feature-based models and demonstrating practical attack methods across multiple file types.

Contribution

It introduces a novel explainability-guided approach for creating backdoor triggers in malware classifiers, applicable across diverse datasets and model types, with practical watermarking implementations.

Findings

Effective backdoor attacks demonstrated on Windows PE, PDFs, and Android classifiers

Explainability techniques improve trigger selection for model-agnostic attacks

Defensive strategies face significant challenges against these sophisticated attacks

Abstract

Training pipelines for machine learning (ML) based malware classification often rely on crowdsourced threat feeds, exposing a natural attack injection point. In this paper, we study the susceptibility of feature-based ML malware classifiers to backdoor poisoning attacks, specifically focusing on challenging "clean label" attacks where attackers do not control the sample labeling process. We propose the use of techniques from explainable machine learning to guide the selection of relevant features and values to create effective backdoor triggers in a model-agnostic fashion. Using multiple reference datasets for malware classification, including Windows PE files, PDFs, and Android applications, we demonstrate effective attacks against a diverse set of machine learning models and evaluate the effect of various constraints imposed on the attacker. To demonstrate the feasibility of our backdoor attacks in practice, we create a watermarking utility for Windows PE files that preserves the binary's functionality, and we leverage similar behavior-preserving alteration methodologies for Android and PDF files. Finally, we experiment with potential defensive strategies and show the difficulties of completely defending against these attacks, especially when the attacks blend in with the legitimate sample distribution.