ML Attack Models: Adversarial Attacks and Data Poisoning Attacks

TL;DR

This paper reviews the vulnerabilities of machine learning models to adversarial and data poisoning attacks, highlighting their threat to the robustness of ML systems in security-sensitive applications.

Contribution

It provides a comprehensive overview of adversarial and data poisoning attacks, emphasizing their impact on the security and robustness of ML models.

Findings

ML models are vulnerable to small perturbations and data poisoning.

Adversarial attacks can fool classifiers with minimal modifications.

Data poisoning can degrade model performance significantly.

Abstract

Many state-of-the-art ML models have outperformed humans in various tasks such as image classification. With such outstanding performance, ML models are widely used today. However, the existence of adversarial attacks and data poisoning attacks really questions the robustness of ML models. For instance, Engstrom et al. demonstrated that state-of-the-art image classifiers could be easily fooled by a small rotation on an arbitrary image. As ML systems are being increasingly integrated into safety and security-sensitive applications, adversarial attacks and data poisoning attacks pose a considerable threat. This chapter focuses on the two broad and important areas of ML security: adversarial attacks and data poisoning attacks.