Early Methods for Detecting Adversarial Images

TL;DR

This paper introduces three methods to detect adversarial images in machine learning classifiers, highlighting that adversarial examples often emphasize lower-ranked PCA components, and discusses the robustness of these detection techniques.

Contribution

The paper presents novel detection methods for adversarial images, emphasizing PCA-based analysis and robustness against adversarial bypass strategies.

Findings

Adversarial images focus on lower-ranked PCA components.

Detection methods can identify adversarial images effectively.

Adversaries must alter images significantly to bypass detection.

Abstract

Many machine learning classifiers are vulnerable to adversarial perturbations. An adversarial perturbation modifies an input to change a classifier's prediction without causing the input to seem substantially different to human perception. We deploy three methods to detect adversarial images. Adversaries trying to bypass our detectors must make the adversarial image less pathological or they will fail trying. Our best detection method reveals that adversarial images place abnormal emphasis on the lower-ranked principal components from PCA. Other detectors and a colorful saliency map are in an appendix.