Membership Inference Attacks via Adversarial Examples

07/27/2022
by   Hamid Jalalzai, et al.
0

The raise of machine learning and deep learning led to significant improvement in several domains. This change is supported by both the dramatic rise in computation power and the collection of large datasets. Such massive datasets often include personal data which can represent a threat to privacy. Membership inference attacks are a novel direction of research which aims at recovering training data used by a learning algorithm. In this paper, we develop a mean to measure the leakage of training data leveraging a quantity appearing as a proxy of the total variation of a trained model near its training samples. We extend our work by providing a novel defense mechanism. Our contributions are supported by empirical evidence through convincing numerical experiments.

READ FULL TEXT

page 3

page 12

research
03/17/2022

Leveraging Adversarial Examples to Quantify Membership Information Leakage

The use of personal data for training machine learning systems comes wit...
research
05/24/2019

Privacy Risks of Securing Machine Learning Models against Adversarial Examples

The arms race between attacks and defenses for machine learning models h...
research
07/28/2020

Label-Only Membership Inference Attacks

Membership inference attacks are one of the simplest forms of privacy le...
research
07/12/2022

RelaxLoss: Defending Membership Inference Attacks without Losing Utility

As a long-term threat to the privacy of training data, membership infere...
research
12/03/2018

Comprehensive Privacy Analysis of Deep Learning: Stand-alone and Federated Learning under Passive and Active White-box Inference Attacks

Deep neural networks are susceptible to various inference attacks as the...
research
02/02/2022

Parameters or Privacy: A Provable Tradeoff Between Overparameterization and Membership Inference

A surprising phenomenon in modern machine learning is the ability of a h...
research
11/17/2021

Do Not Trust Prediction Scores for Membership Inference Attacks

Membership inference attacks (MIAs) aim to determine whether a specific ...

Please sign up or login with your details

Forgot password? Click here to reset