Label-Only Membership Inference Attacks and Defenses in Semantic Segmentation Models
- 1 March 2022
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Dependable and Secure Computing
- Vol. 20 (2) , 1435-1449
Abstract
Recent research has discovered that deep learning models are vulnerable to membership inference attacks, which can reveal whether a sample is in the training dataset of the victim model or not. Most membership inference attacks rely on confidence scores from the victim model for the attack purpose. However, a few studies indicate that prediction labels of the victim model's output are sufficient for launching successful attacks. Besides the well-studied classification models, segmentation models are also vulnerable to this type of attack. In this paper, for the first time, we propose the label-only membership inference attacks against semantic segmentation models. With a well-designed framework of the attacks, we can achieve a considerably higher successful attacking rate compared to previous work. In addition, we have discussed several possible defense mechanisms to counter such a threat.Keywords
Funding Information
- ARC (LP180101150, DP190100981)
- Australian Research Council
This publication has 32 references indexed in Scilit:
- Membership Inference Attacks Against Machine Learning Models2017
- SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image SegmentationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2017
- Deep Learning with Differential Privacy2016
- Empirical Risk Minimization2016
- Deep Residual Learning for Image Recognition2016
- The Cityscapes Dataset for Semantic Urban Scene Understanding2016
- Attention to Scale: Scale-Aware Semantic Image Segmentation2016
- Fully convolutional networks for semantic segmentation2015
- An introduction to ROC analysisPattern Recognition Letters, 2006
- Our Data, Ourselves: Privacy Via Distributed Noise Generation2006