ALLUDE: A Unified Evaluation System for Configurable Attacks in Differentiable Environments

2026-07-19

AI Summary

Current evaluations of visual adversarial attacks are often conducted under constrained conditions, failing to reflect performance in real-world, complex scenarios. This work proposes the first unified evaluation framework spanning both Linux and Windows platforms, featuring end-to-end differentiable rendering and enabling large-scale combinatorial configuration of multidimensional environmental parameters—including weather, illumination, object properties, and camera trajectories—via Latin hypercube sampling. Integrating multiple detection models, the system samples and validates across 5,400 distinct configurations, revealing for the first time the significant performance degradation of prominent attacks such as CAMOU, RAUCA, and FCA under dynamic, realistic conditions. This study bridges a critical gap in adversarial robustness evaluation by enabling high-fidelity, diverse scenario testing.

Abstract

Adversarial attacks against vision models like object detectors are often evaluated under limited conditions, leaving their performance under-characterized. Bridging simulation and differentiable rendering enables more robust, end-to-end evaluation of these adversarial attacks, yet there is no easy-to-use, unified system that offers a rich set of customizable configurations for adversarial attacks across multiple scenes, objects, environmental and lighting conditions, and camera trajectories. We present ALLUDE, which addresses these gaps, offering first-of-its-kind evaluation capabilities across Linux and Windows. We comprehensively demonstrate ALLUDE's evaluation breadth through a two-pronged strategy: (1) using Latin Hypercube Sampling, we draw a representative subset from 5,400 configurations spanning 10 scene-object pairs, 9 weather conditions, 4 optimizers, 5 camera trajectories, and 3 detection models; (2) we stress-test existing attacks (CAMOU, RAUCA, FCA) under diverse weather conditions and continuous camera trajectories, revealing degradation of attack success across every attack, exposing evaluation gaps in prior work. Through ALLUDE's end-to-end differentiable rendering, adversarial attacks can be optimized against shifting real-world deployment conditions. Our cross-platform code is open source.

Problem

• adversarial attacks
• evaluation system
• differentiable rendering
• configurable conditions
• vision models

Innovation

• differentiable rendering
• adversarial attack evaluation
• configurable simulation
• cross-platform framework
• Latin Hypercube Sampling