Membership Inference Attacks Against Object Detection Models
share
Machine learning models can leak information about the dataset they trained. In this paper, we present the first membership inference attack against black-boxed object detection models that determines whether given data records are used in training. To attack the object detection model, we devise a noble method named as Canvas Method, drawing predicted bounding boxes on an empty image for attack model input. In experiments, we successfully reveal the membership status of privately sensitive data trained by one-stage and two-stage detection models. We then propose defense strategies and also conduct a transfer attack between models and datasets. Our results show that object detection models are vulnerable to inference attacks as other models.
READ FULL TEXT