Unsupervised Multi-Criteria Adversarial Detection in Deep Image Retrieval

TL;DR

This paper introduces an unsupervised method for detecting adversarial attacks in deep hashing-based image retrieval systems by analyzing behaviors in the hamming space, improving detection rates with minimal overhead.

Contribution

It proposes a novel unsupervised detection scheme using three criteria in the hamming space, tailored for deep hashing image retrieval, addressing a gap in existing defense strategies.

Findings

Achieves 2-23% higher detection rates across four datasets.

Operates with minimal computational overhead for real-time applications.

Effectively defends against both untargeted and targeted adversarial attacks.

Abstract

The vulnerability in the algorithm supply chain of deep learning has imposed new challenges to image retrieval systems in the downstream. Among a variety of techniques, deep hashing is gaining popularity. As it inherits the algorithmic backend from deep learning, a handful of attacks are recently proposed to disrupt normal image retrieval. Unfortunately, the defense strategies in softmax classification are not readily available to be applied in the image retrieval domain. In this paper, we propose an efficient and unsupervised scheme to identify unique adversarial behaviors in the hamming space. In particular, we design three criteria from the perspectives of hamming distance, quantization loss and denoising to defend against both untargeted and targeted attacks, which collectively limit the adversarial space. The extensive experiments on four datasets demonstrate 2-23% improvements of detection rates with minimum computational overhead for real-time image queries.