TIJO: Trigger Inversion with Joint Optimization for Defending Multimodal Backdoored Models

TL;DR

TIJO is a novel joint optimization method that effectively reverses backdoor triggers in multimodal models by operating in the object detection feature space, significantly improving detection accuracy on the TrojVQA benchmark.

Contribution

This paper introduces TIJO, the first joint optimization approach for backdoor trigger inversion in multimodal models, addressing challenges posed by disconnected visual pipelines.

Findings

TIJO achieves an AUC of 0.92 on TrojVQA, surpassing previous unimodal methods.

TIJO improves detection performance on both multimodal and unimodal backdoors.

Ablation studies highlight the importance of overlaying inverted triggers on all visual features.

Abstract

We present a Multimodal Backdoor Defense technique TIJO (Trigger Inversion using Joint Optimization). Recent work arXiv:2112.07668 has demonstrated successful backdoor attacks on multimodal models for the Visual Question Answering task. Their dual-key backdoor trigger is split across two modalities (image and text), such that the backdoor is activated if and only if the trigger is present in both modalities. We propose TIJO that defends against dual-key attacks through a joint optimization that reverse-engineers the trigger in both the image and text modalities. This joint optimization is challenging in multimodal models due to the disconnected nature of the visual pipeline which consists of an offline feature extractor, whose output is then fused with the text using a fusion module. The key insight enabling the joint optimization in TIJO is that the trigger inversion needs to be carried out in the object detection box feature space as opposed to the pixel space. We demonstrate the effectiveness of our method on the TrojVQA benchmark, where TIJO improves upon the state-of-the-art unimodal methods from an AUC of 0.6 to 0.92 on multimodal dual-key backdoors. Furthermore, our method also improves upon the unimodal baselines on unimodal backdoors. We present ablation studies and qualitative results to provide insights into our algorithm such as the critical importance of overlaying the inverted feature triggers on all visual features during trigger inversion. The prototype implementation of TIJO is available at https://github.com/SRI-CSL/TIJO.