Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Can You Trust What You See? Alpha Channel No-Box Attacks on Video Object Detection

Ariana Yi, Ce Zhou, Liyang Xiao, Qiben Yan

TL;DR

α-Cloak reveals a previously overlooked vulnerability in video perception by embedding adversarial payloads in the alpha channel of RGBA videos. The method fuses a benign human-visible stream with a malicious target in a frame-wise alpha-blending framework, producing videos that appear normal to humans but fool RGB-based detectors after alpha removal. The approach is no-box, architecture- and query-free, and demonstrates 100% success across multiple object detectors, a vision-language model, and a multimodal LLM, across diverse datasets and formats. This work underscores the need for defenses that explicitly address alpha-channel handling to ensure consistency between human and machine perception in video pipelines.

Abstract

As object detection models are increasingly deployed in cyber-physical systems such as autonomous vehicles (AVs) and surveillance platforms, ensuring their security against adversarial threats is essential. While prior work has explored adversarial attacks in the image domain, those attacks in the video domain remain largely unexamined, especially in the no-box setting. In this paper, we present α-Cloak, the first no-box adversarial attack on object detectors that operates entirely through the alpha channel of RGBA videos. α-Cloak exploits the alpha channel to fuse a malicious target video with a benign video, resulting in a fused video that appears innocuous to human viewers but consistently fools object detectors. Our attack requires no access to model architecture, parameters, or outputs, and introduces no perceptible artifacts. We systematically study the support for alpha channels across common video formats and playback applications, and design a fusion algorithm that ensures visual stealth and compatibility. We evaluate α-Cloak on five state-of-the-art object detectors, a vision-language model, and a multi-modal large language model (Gemini-2.0-Flash), demonstrating a 100% attack success rate across all scenarios. Our findings reveal a previously unexplored vulnerability in video-based perception systems, highlighting the urgent need for defenses that account for the alpha channel in adversarial settings.

Can You Trust What You See? Alpha Channel No-Box Attacks on Video Object Detection

TL;DR

α-Cloak reveals a previously overlooked vulnerability in video perception by embedding adversarial payloads in the alpha channel of RGBA videos. The method fuses a benign human-visible stream with a malicious target in a frame-wise alpha-blending framework, producing videos that appear normal to humans but fool RGB-based detectors after alpha removal. The approach is no-box, architecture- and query-free, and demonstrates 100% success across multiple object detectors, a vision-language model, and a multimodal LLM, across diverse datasets and formats. This work underscores the need for defenses that explicitly address alpha-channel handling to ensure consistency between human and machine perception in video pipelines.

Abstract

As object detection models are increasingly deployed in cyber-physical systems such as autonomous vehicles (AVs) and surveillance platforms, ensuring their security against adversarial threats is essential. While prior work has explored adversarial attacks in the image domain, those attacks in the video domain remain largely unexamined, especially in the no-box setting. In this paper, we present α-Cloak, the first no-box adversarial attack on object detectors that operates entirely through the alpha channel of RGBA videos. α-Cloak exploits the alpha channel to fuse a malicious target video with a benign video, resulting in a fused video that appears innocuous to human viewers but consistently fools object detectors. Our attack requires no access to model architecture, parameters, or outputs, and introduces no perceptible artifacts. We systematically study the support for alpha channels across common video formats and playback applications, and design a fusion algorithm that ensures visual stealth and compatibility. We evaluate α-Cloak on five state-of-the-art object detectors, a vision-language model, and a multi-modal large language model (Gemini-2.0-Flash), demonstrating a 100% attack success rate across all scenarios. Our findings reveal a previously unexplored vulnerability in video-based perception systems, highlighting the urgent need for defenses that account for the alpha channel in adversarial settings.
Paper Structure (28 sections, 7 equations, 4 figures, 6 tables, 2 algorithms)

This paper contains 28 sections, 7 equations, 4 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Background
  3. RGBA Video Format and Alpha Channel
  4. Background Colors of Video Player Applications
  5. Alpha Channel in Video File Types
  6. RGBA Video-Based Object Detectors Processing Pipeline
  7. Threat Model
  8. $\alpha$-Cloak Attack Design
  9. Attack Overview
  10. Video Frame Preprocessing
  11. Video Frames Combination
  12. Evaluation
  13. Experimental Setup
  14. Experimental Procedure.
  15. Target Attack Models.
  16. ...and 13 more sections

Figures (4)

  • Figure 1: An example of alpha channel blending illustrates how the alpha value interacts with the RGB channels and the background color to control pixel transparency and the final color seen in video display applications.
  • Figure 2: Attack scenario
  • Figure 3: Overview of the $\alpha$-Cloak pipeline
  • Figure 4: Evaluation results across different detection models. The x-axis represents each attacked video; the y-axis is the VLS (%). Red: $V_{\text{FAKE}}$; blue: $V_{\text{TRUE}}$.