Who's Afraid of Adversarial Queries? The Impact of Image Modifications on Content-based Image Retrieval

TL;DR

This paper introduces PIRE, an unsupervised neural perturbation method that effectively disrupts content-based image retrieval systems, highlighting new challenges and considerations for future multimedia research.

Contribution

The paper presents PIRE, a novel unsupervised neural approach to generate adversarial queries that block neural-feature-based CBIR without requiring labeled data.

Findings

PIRE effectively blocks neural-feature-based CBIR.

It performs comparably or better than keypoint modification methods.

Practical considerations include image quality and background leakage.

Abstract

An adversarial query is an image that has been modified to disrupt content-based image retrieval (CBIR) while appearing nearly untouched to the human eye. This paper presents an analysis of adversarial queries for CBIR based on neural, local, and global features. We introduce an innovative neural image perturbation approach, called Perturbations for Image Retrieval Error (PIRE), that is capable of blocking neural-feature-based CBIR. PIRE differs significantly from existing approaches that create images adversarial with respect to CNN classifiers because it is unsupervised, i.e., it needs no labelled data from the data set to which it is applied. Our experimental analysis demonstrates the surprising effectiveness of PIRE in blocking CBIR, and also covers aspects of PIRE that must be taken into account in practical settings, including saving images, image quality and leaking adversarial queries into the background collection. Our experiments also compare PIRE (a neural approach) with existing keypoint removal and injection approaches (which modify local features). Finally, we discuss the challenges that face multimedia researchers in the future study of adversarial queries.