Label-Only Membership Inference Attacks

Authors: Christopher A. Choquette-Choo, Florian Tramer, Nicholas Carlini, Nicolas Papernot

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments show that training with differential privacy or strong ℓ2 regularization are the only current defenses that meaningfully decrease leakage of private information, even for points that are outliers of the training distribution.
Researcher Affiliation Collaboration 1University of Toronto and Vector Institute 2Stanford University 3Google. Correspondence to: Christopher A. Choquette-Choo <choquette.christopher@gmail.com>.
Pseudocode No The paper describes its methods in prose, but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Our code is available at https://github.com/cchoquette/ membership-inference.
Open Datasets Yes We evaluate our attacks on 8 datasets used by the canonical work of Shokri et al. (2016). These include 3 computer vision tasks3... and 4 non-computer-vision tasks4... 3MNIST, CIFAR-10, and CIFAR-100: https://www.tensorflow.org/api_docs/python/tf/keras/datasets 4Adult Dataset: http://archive.ics.uci.edu/ml/ datasets/Adult Texas-100, Purchase-100, and Locations datasets: https://github.com/privacytrustlab/datasets
Dataset Splits No The paper mentions training data, held-out data, and test accuracy but does not specify explicit train/validation/test dataset splits with percentages or counts, nor does it explicitly mention a validation set.
Hardware Specification No The paper does not provide specific details regarding the hardware (e.g., GPU models, CPU types, or cloud instance specifications) used for running the experiments.
Software Dependencies No The paper mentions 'TensorFlow' in a footnote related to datasets, but it does not specify version numbers for TensorFlow or any other software dependencies, libraries, or programming languages used.
Experiment Setup Yes We train target models with data augmentation similar to 3.3 and focus on translations as they are most common in computer vision. We use a simple pipeline where all translations of each image is evaluated in a training epoch. ... (non-random) weight decay of magnitude 0.0005.