A Survey of Privacy Attacks in Machine Learning (2020-07-15T00:00:00.000000Z)

TL;DR

An attack taxonomy is proposed, together with a threat model that allows the categorization of different attacks based on the adversarial knowledge, and the assets under attack, and a discussion of the open problems and future directions identified during the analysis.

Abstract

As machine learning becomes more widely used, the need to study its implications in security and privacy becomes more urgent. Although the body of work in privacy has been steadily growing over the past few years, research on the privacy aspects of machine learning has received less focus than the security aspects. Our contribution in this research is an analysis of more than 45 papers related to privacy attacks against machine learning that have been published during the past seven years. We propose an attack taxonomy, together with a threat model that allows the categorization of different attacks based on the adversarial knowledge, and the assets under attack. An initial exploration of the causes of privacy leaks is presented, as well as a detailed analysis of the different attacks. Finally, we present an overview of the most commonly proposed defenses and a discussion of the open problems and future directions identified during our analysis.

Authors

M. Rigaki

2 papers

S. García

1 papers

TL;DR

Abstract

Authors

References158 items

InverseNet: Augmenting Model Extraction Attacks with Training Data Inversion

A Taxonomy of Attacks on Federated Learning

ML-Doctor: Holistic Risk Assessment of Inference Attacks Against Machine Learning Models

Extracting Training Data from Large Language Models

Netflix

Black-Box Ripper: Copying black-box models using generative evolutionary algorithms

Membership Leakage in Label-Only Exposures

Label-Only Membership Inference Attacks

TUDataset: A collection of benchmark datasets for learning with graphs

Privacy Risks of General-Purpose Language Models

Diabetic Retinopathy Detection

ActiveThief: Model Extraction Using Active Learning and Unannotated Public Data

Information Leakage in Embedding Models

Segmentations-Leak: Membership Inference Attacks and Defenses in Semantic Image Segmentation

Model inversion attacks against collaborative inference

The Secret Revealer: Generative Model-Inversion Attacks Against Deep Neural Networks

Neural Network Inversion in Adversarial Setting via Background Knowledge Alignment

Thieves on Sesame Street! Model Extraction of BERT-based APIs

MemGuard: Defending against Black-Box Membership Inference Attacks via Adversarial Examples

BDPL: A Boundary Differentially Private Layer Against Machine Learning Model Extraction Attacks

GAN-Leaks: A Taxonomy of Membership Inference Attacks against Generative Models

GAN-Leaks: A Taxonomy of Membership Inference Attacks against GANs

High Accuracy and High Fidelity Extraction of Neural Networks

Federated Learning: Challenges, Methods, and Future Directions

The security of machine learning in an adversarial setting: A survey

Privacy Risks of Explaining Machine Learning Models

Stolen Memories: Leveraging Model Memorization for Calibrated White-Box Membership Inference

Prediction Poisoning: Towards Defenses Against DNN Model Stealing Attacks

Monte Carlo and Reconstruction Membership Inference Attacks against Generative Models

Deep Leakage from Gradients

Overlearning Reveals Sensitive Attributes

White-box vs Black-box: Bayes Optimal Strategies for Membership Inference

Privacy Risks of Securing Machine Learning Models against Adversarial Examples

S4L: Self-Supervised Semi-Supervised Learning

MixMatch: A Holistic Approach to Semi-Supervised Learning

Membership Inference Attacks on Sequence-to-Sequence Models: Is My Data In Your Machine Translation System?

Updates-Leak: Data Set Inference and Reconstruction Attacks in Online Learning

Evaluating Differentially Private Machine Learning in Practice

Demystifying Membership Inference Attacks in Machine Learning as a Service

Knockoff Nets: Stealing Functionality of Black-Box Models

Comprehensive Privacy Analysis of Deep Learning: Passive and Active White-box Inference Attacks against Centralized and Federated Learning

Beyond Inferring Class Representatives: User-Level Privacy Leakage From Federated Learning

Exploring Connections Between Active Learning and Model Extraction

Towards Adversarial Malware Detection

Auditing Data Provenance in Text-Generation Models

Property Inference Attacks on Fully Connected Neural Networks using Permutation Invariant Representations

Numerical Optimization

Model Reconstruction from Model Explanations

Algorithms that remember: model inversion attacks and data protection law

Copycat CNN: Stealing Knowledge by Persuading Confession with Random Non-Labeled Data

ML-Leaks: Model and Data Independent Membership Inference Attacks and Defenses on Machine Learning Models

AttriGuard: A Practical Defense Against Attribute Inference Attacks via Adversarial Machine Learning

Exploiting Unintended Feature Leakage in Collaborative Learning

PRADA: Protecting Against DNN Model Stealing Attacks

SoK: Security and Privacy in Machine Learning

Privacy-Preserving Machine Learning: Threats and Solutions

The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks

Stealing Hyperparameters in Machine Learning

Understanding Membership Inferences on Well-Generalized Learning Models

Machine Learning with Membership Privacy using Adversarial Regularization

Wild Patterns: Ten Years After the Rise of Adversarial Machine Learning

LightGBM: A Highly Efficient Gradient Boosting Decision Tree

Towards Reverse-Engineering Black-Box Neural Networks

Oblivious Neural Network Predictions via MiniONN Transformations

Learning Differentially Private Recurrent Language Models

Privacy Risk in Machine Learning: Analyzing the Connection to Overfitting

walk2friends: Inferring Social Links from Mobility Profiles

Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms

Active Learning for Convolutional Neural Networks: A Core-Set Approach

Model Inversion Attacks for Prediction Systems: Without Knowledge of Non-Sensitive Attributes

Privacy-Preserving Deep Learning: Revisited and Enhanced

Towards Deep Learning Models Resistant to Adversarial Attacks

LOGAN: Membership Inference Attacks Against Generative Models

ChestX-Ray8: Hospital-Scale Chest X-Ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases

Neurosurgeon: Collaborative Intelligence Between the Cloud and Mobile Edge

Deep Models Under the GAN: Information Leakage from Collaborative Deep Learning