HAWKEYE: Adversarial Example Detector for Deep Neural Networks

TL;DR

HAWKEYE is a neural network-based adversarial example detector that analyzes DNN output variations using quantized references, significantly reducing false positives while maintaining high detection rates.

Contribution

The paper introduces HAWKEYE, a novel AE detection method that leverages output analysis with quantized references and cascading detectors for improved accuracy.

Findings

Reduces false positive rate in AE detection.

Maintains high detection rate for adversarial examples.

Cascading detectors enhance overall detection performance.

Abstract

Adversarial examples (AEs) are images that can mislead deep neural network (DNN) classifiers via introducing slight perturbations into original images. Recent work has shown that detecting AEs can be more effective against AEs than preventing them from being generated. However, the state-of-the-art AE detection still shows a high false positive rate, thereby rejecting a considerable amount of normal images. To address this issue, we propose HAWKEYE, which is a separate neural network that analyzes the output layer of the DNN, and detects AEs. HAWKEYE's AE detector utilizes a quantized version of an input image as a reference, and is trained to distinguish the variation characteristics of the DNN output on an input image from the DNN output on its reference image. We also show that cascading our AE detectors that are trained for different quantization step sizes can drastically reduce a false positive rate, while keeping a detection rate high.