Stealthy Targeted Backdoor Attacks against Image Captioning

TL;DR

This paper introduces a stealthy backdoor attack method on image captioning models that manipulates object recognition to produce targeted captions, achieving high success rates without affecting normal performance.

Contribution

The paper proposes a novel, stealthier backdoor attack technique that leverages universal perturbations and object placement to manipulate image captions.

Findings

High attack success rate achieved

Backdoor samples are indistinguishable from clean samples

Method bypasses existing defenses

Abstract

In recent years, there has been an explosive growth in multimodal learning. Image captioning, a classical multimodal task, has demonstrated promising applications and attracted extensive research attention. However, recent studies have shown that image caption models are vulnerable to some security threats such as backdoor attacks. Existing backdoor attacks against image captioning typically pair a trigger either with a predefined sentence or a single word as the targeted output, yet they are unrelated to the image content, making them easily noticeable as anomalies by humans. In this paper, we present a novel method to craft targeted backdoor attacks against image caption models, which are designed to be stealthier than prior attacks. Specifically, our method first learns a special trigger by leveraging universal perturbation techniques for object detection, then places the learned trigger in the center of some specific source object and modifies the corresponding object name in the output caption to a predefined target name. During the prediction phase, the caption produced by the backdoored model for input images with the trigger can accurately convey the semantic information of the rest of the whole image, while incorrectly recognizing the source object as the predefined target. Extensive experiments demonstrate that our approach can achieve a high attack success rate while having a negligible impact on model clean performance. In addition, we show our method is stealthy in that the produced backdoor samples are indistinguishable from clean samples in both image and text domains, which can successfully bypass existing backdoor defenses, highlighting the need for better defensive mechanisms against such stealthy backdoor attacks.