Abstract:As object detection becomes integral to many safety-critical applications, understanding its vulnerabilities is essential. Backdoor attacks, in particular, pose a significant threat by implanting hidden backdoor in a victim model, which adversaries can later exploit to trigger malicious behaviors during inference. However, current backdoor techniques are limited to static scenarios where attackers must define a malicious objective before training, locking the attack into a predetermined action without inference-time adaptability. Given the expressive output space in object detection, including object existence detection, bounding box estimation, and object classification, the feasibility of implanting a backdoor that provides inference-time control with a high degree of freedom remains unexplored. This paper introduces AnywhereDoor, a flexible backdoor attack tailored for object detection. Once implanted, AnywhereDoor enables adversaries to specify different attack types (object vanishing, fabrication, or misclassification) and configurations (untargeted or targeted with specific classes) to dynamically control detection behavior. This flexibility is achieved through three key innovations: (i) objective disentanglement to support a broader range of attack combinations well beyond what existing methods allow; (ii) trigger mosaicking to ensure backdoor activations are robust, even against those object detectors that extract localized regions from the input image for recognition; and (iii) strategic batching to address object-level data imbalances that otherwise hinders a balanced manipulation. Extensive experiments demonstrate that AnywhereDoor provides attackers with a high degree of control, achieving an attack success rate improvement of nearly 80% compared to adaptations of existing methods for such flexible control.

Door Detection without Apriori Color Knowledge

Window Detection in Facades Using Heatmap Fusion

A Self-Organizing Neural Scheme for Door Detection in Different Environments

Door State Recognition Method for Wall Reconstruction from Scanned Scene in Point Clouds

Real-time Door Detection for Indoor Autonomous Vehicle

Foreground Gating and Background Refining Network for Surveillance Object Detection

Intelligent Video Surveillance for Checking Attendance of Traffic Controllers in Level Crossing

Enhancing Door-Status Detection for Autonomous Mobile Robots during Environment-Specific Operational Use

An intelligent automated door control system based on a smart camera

Real-time 2D-3D door detection and state classification on a low-power device

Detection-Free Framework for Cabinet Switch State Recognition

RGBD-based Parameter Extraction for Door Opening Tasks with Human Assists in Nuclear Rescue

Detection and Measurement of Opening and Closing Automatic Sliding Glass Doors

A Fire-Alarming Method Based on Video Processing

Detecting More Objects Indoors: a Depth-guided Detector for Mobile Robots

A Real World Detection System - Combining Color, Shape and Appearance to Enable Real-time Road Sign Detection.

Learning Semantic Keypoint Representations for Door Opening Manipulation

DQ-HorizonNet: Enhancing Door Detection Accuracy in Panoramic Images Via Dynamic Quantization

AnywhereDoor: Multi-Target Backdoor Attacks on Object Detection

Development and Adaptation of Robotic Vision in the Real-World: the Challenge of Door Detection

Real-Time Detection of Face Abnormal Occlusion Based on the Indoor Patrol Car