NoPain: No-box Point Cloud Attack via Optimal Transport Singular Boundary

2025-02-27

AI Summary

Existing point-cloud adversarial attack methods rely on surrogate models and gradient-based iterative optimization, suffering from overfitting and poor transferability. This paper proposes a model-free, iteration-free, and query-free (no-box) adversarial generation paradigm. To our knowledge, it is the first to introduce optimal transport theory into point-cloud attacks: by modeling the optimal transport mapping over the point-cloud manifold, the method identifies non-differentiable intrinsic singular boundaries in feature space and generates gradient-free adversarial perturbations via manifold-geometric sampling. The approach requires no training, neither gradients nor surrogate models. It significantly improves cross-architecture transferability (average +12.7%), accelerates inference by 8.3×, and maintains strong robustness against mainstream point-cloud defenses.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Adversarial Search

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Adversarial attacks exploit the vulnerability of deep models against adversarial samples. Existing point cloud attackers are tailored to specific models, iteratively optimizing perturbations based on gradients in either a white-box or black-box setting. Despite their promising attack performance, they often struggle to produce transferable adversarial samples due to overfitting the specific parameters of surrogate models. To overcome this issue, we shift our focus to the data distribution itself and introduce a novel approach named extbf{NoPain}, which employs optimal transport (OT) to identify the inherent singular boundaries of the data manifold for cross-network point cloud attacks. Specifically, we first calculate the OT mapping from noise to the target feature space, then identify singular boundaries by locating non-differentiable positions. Finally, we sample along singular boundaries to generate adversarial point clouds. Once the singular boundaries are determined, NoPain can efficiently produce adversarial samples without the need of iterative updates or guidance from the surrogate classifiers. Extensive experiments demonstrate that the proposed end-to-end method outperforms baseline approaches in terms of both transferability and efficiency, while also maintaining notable advantages even against defense strategies. The source code will be publicly available.

Problem

• Overcoming overfitting in adversarial point cloud attacks
• Generating transferable adversarial samples without iterative updates
• Enhancing attack efficiency and transferability across networks

Innovation

• Uses optimal transport for singular boundary identification
• Generates adversarial samples without iterative updates
• Enhances transferability and efficiency in attacks