Vision-Based Adversarial Sample Detection for Road Sign Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional motion planning and control systems for autonomous vehicles do not accurately account for differences in vehicle types, leading to potential inaccuracies and inefficiencies, and deep learning-based vision systems are vulnerable to adversarial samples that can misinterpret road signs, posing safety risks.
Innovation Solution
A system that applies image transformations to detect adversarial samples by calculating a similarity metric between original and transformed images, using object detection models to identify potential adversarial inputs and activate failsafe mechanisms if inconsistencies are detected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning based computer vision models are used for object detection, then detection accuracy is improved, but vulnerability to adversarial samples increases
Solution Approach 1:
The system applies image transformations to the input image before object detection to generate transformed images. This preliminary action creates multiple variations of the image that are used to detect adversarial samples by comparing detection results across transformations, thereby preventing adversarial attacks from compromising the detection accuracy of deep learning models.
Solution Approach 2:
The system introduces an intermediary detection mechanism that acts as a mediator between the input image and the deep learning model. By computing similarity metrics between detection results of original and transformed images, this intermediary layer identifies adversarial samples before they can mislead the object detection model, thus protecting system reliability while maintaining detection accuracy.
2Reliability
If image transformations are applied to detect adversarial samples, then system reliability is improved, but computational complexity increases
Solution Approach 1:
The system applies image transformations selectively rather than exhaustively. By choosing a limited set of transformations and applying them only when needed for adversarial detection, the system achieves sufficient reliability improvement without incurring excessive computational complexity. The transformation process is optimized to balance detection effectiveness with computational resources.
Data Source
AI summary
In one embodiment, a system receives a first image captured by a capturing device of an ADV. The system applies an image transformation to the first image to generate a second image. The system applies an object detection model to the first and the second images to generate a first and a second output, respectively. The system calculates a similarity metric between the first and the second output. The system detects the first image as an adversarial sample if a temporal variation of the similarity metric between the first image and a prior image is above a threshold.