6DAttack: Backdoor Attacks in the 6DoF Pose Estimation

TL;DR

This paper introduces 6DAttack, a novel backdoor attack framework targeting 6DoF pose estimation models using 3D triggers, demonstrating high success rates and exposing security vulnerabilities in critical applications.

Contribution

We propose the first backdoor attack method specifically designed for 6DoF pose estimation, addressing continuous parameter control and evaluating its effectiveness across multiple models and datasets.

Findings

High attack success rates (up to 100%) without degrading clean performance.

Backdoored models maintain high accuracy on clean data.

Existing defenses are ineffective against 6DoF backdoor attacks.

Abstract

Deep learning advances have enabled accurate six-degree-of-freedom (6DoF) object pose estimation, widely used in robotics, AR/VR, and autonomous systems. However, backdoor attacks pose significant security risks. While most research focuses on 2D vision, 6DoF pose estimation remains largely unexplored. Unlike traditional backdoors that only change classes, 6DoF attacks must control continuous parameters like translation and rotation, rendering 2D methods inapplicable. We propose 6DAttack, a framework using 3D object triggers to induce controlled erroneous poses while maintaining normal behavior. Evaluations on PVNet, DenseFusion, and PoseDiffusion across LINEMOD, YCB-Video, and CO3D show high attack success rates (ASRs) without compromising clean performance. Backdoored models achieve up to 100% clean ADD accuracy and 100% ASR, with triggered samples reaching 97.70% ADD-P. Furthermore, a representative defense remains ineffective. Our findings reveal a serious, underexplored threat to 6DoF pose estimation.