LatentStealth: Unnoticeable and Efficient Adversarial Attacks on Expressive Human Pose and Shape Estimation

TL;DR

LatentStealth introduces an unnoticeable, efficient adversarial attack method on human pose estimation models by manipulating structured latent representations, exposing security vulnerabilities with minimal perceptibility.

Contribution

The paper presents a novel latent-space adversarial attack technique that maintains high imperceptibility and low computational cost, revealing vulnerabilities in current human pose estimation systems.

Findings

LatentStealth achieves high attack success with minimal perceptible changes.

Experiments on 3DPW and UBody datasets demonstrate its effectiveness.

The method exposes critical security vulnerabilities in EHPS models.

Abstract

Expressive human pose and shape estimation (EHPS) plays a central role in digital human generation, particularly in live-streaming applications. However, most existing EHPS models focus primarily on minimizing estimation errors, with limited attention on potential security vulnerabilities, such as generating inappropriate content, violent actions, or racially offensive gestures and expressions. Current adversarial attacks on EHPS models often generate visually conspicuous perturbations, limiting their practicality and ability to expose real-world security threats. To address this limitation, we propose an unnoticeable adversarial method, termed \textbf{LatentStealth}, specifically tailored for EHPS models. The key idea is to exploit the structured latent representations of natural images as the medium for crafting perturbations. Instead of injecting noise directly into the pixel space, our method projects inputs into the latent space, where adversarial patterns are generated and progressively refined along optimized directions. This latent-space manipulation enables the attack to maintain high imperceptibility while preserving its effectiveness. Furthermore, as the optimization process is guided by only a small number of model output queries, the framework achieves competitive attack performance with low computational overhead, making it both practical and efficient for real-world scenarios. Extensive experiments on the 3DPW and UBody datasets demonstrate the superiority of LatentStealth, revealing critical vulnerabilities in current systems. These findings highlight the urgent need to address and mitigate security risks in digital human generation technologies.