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

By Jinyuan Jia, Neil Zhenqiang Gong. 2018

195 citations recorded by the source corpus.

Users in various web and mobile applications are vulnerable to attribute inference attacks, in which an attacker leverages a machine learning classifier to infer a target user's private attributes (e.g., location, sexual orientation, political view) from its public data (e.g., rating scores, page likes). Existing defenses leverage game theory or heuristics based on correlations between the public data and attributes. These defenses are not practical. Specifically, game-theoretic defenses require solving intractable optimization problems, while correlation-based defenses incur large utility loss of users' public data. In this paper, we present AttriGuard, a practical defense against attribute inference attacks. AttriGuard is computationally tractable and has small utility loss. Our AttriGuard works in two phases. Suppose we aim to protect a user's private attribute. In Phase I, for each value of the attribute, we find a minimum noise such that if we add the noise to the user's public data, then the attacker's classifier is very likely to infer the attribute value for the user. We find the minimum noise via adapting existing evasion attacks in adversarial machine learning. In Phase II, we sample one attribute value according to a certain probability distribution and add the corresponding noise found in Phase I to the user's public data. We formulate finding the probability distribution as solving a constrained convex optimization problem. We extensively evaluate AttriGuard and compare it with existing methods using a real-world dataset. Our results show that AttriGuard substantially outperforms existing methods. Our work is the first one that shows evasion attacks can be used as defensive techniques for privacy protection.

Subfields: Artificial Intelligence, Safety Research

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety Research

Citations in the corpus

Attacking Graph-based Classification via Manipulating the Graph Structure (2019, background): “More-over, when an attacker uses graph-based classiﬁcation meth-ods as attribute inference attacks to compromise user privacy, evasion to them can be used as a privacy protection mechanism [33].”
Poisoning Attacks to Graph-Based Recommender Systems (2018, background): “Jia and Gong [15] recently proposed a practical defense against attribute inference attacks via adversarial machine learning. key idea is to add carefully crafted noise to a user’s rating behavior data such that the attacker’s classifier is very likely to make incorrect predictions.”
LOGAN: Membership Inference Attacks Against Generative Models (2017, methodology): “ator of a Generative Adversarial Network under differential privacy. The resulting model is then used to generate synthetic subjects based on the population of clinical trial data. Finally, Jia et al.[28] use adversarial machine learning to defend against attribute inference attacks, in the setting ”
Machine Learning with Membership Privacy using Adversarial Regularization (2018, methodology): “We model this optimization as a min-max privacy game between the defense mechanism and the inference attack, similar to privacy games in other settings [2, 24, 27, 34, 46].”
Vertical Federated Learning: Concepts, Advances, and Challenges (2022, background): “Adversarial training [138, 118] and mutual information regularization [118] were proposed to safeguard sensitive attributes of training samples.”
Neural Network Inversion in Adversarial Setting via Background Knowledge Alignment (2019, background): “Researchers have strongly suggested that ML models suffer from various privacy threats to its training data [3, 22, 31, 58, 65].”
MemGuard: Defending against Black-Box Membership Inference Attacks via Adversarial Examples (2019, background): “Jia and Gong proposed AttriGuard [26], which leverages adversarial examples to defend against attribute inference attacks.”
When Machine Learning Meets Privacy (2020, background): “[65] proposed a two-phase framework called AttriGuard to defend against attribute inference attacks launched by a classifier.”
A Survey of Privacy Attacks in Machine Learning (2020, background): “(3) The model consumers , that use the services that the model owner exposes, usually via some sort of programming or user interface.”
ML-Leaks: Model and Data Independent Membership Inference Attacks and Defenses on Machine Learning Models (2018, background): “On the other hand, researchers have recently shown that adversarial examples can also help to protect users’ privacy in online social networks [22], [31], [46].”
Olympus: Sensor Privacy through Utility Aware Obfuscation (2018, unknown): “AttriGuard [25] uses adversarial perturbations to defend against an attacker that attempts to infer private attributes from the data.”
Privacy Risks of General-Purpose Language Models (2020, background): “Although previous PPM-based defenses have studied several efﬁcient approaches to solve the minimax game when z takes discrete values and D θ is a combinatorial function [35], [57], our PPM is required to work on the real-valued embeddings with D θ implemented as neural networks.”
Adversarial Neural Network Inversion via Auxiliary Knowledge Alignment (2019, background): “ML Privacy Researchers have strongly suggested that ML models suffer from various privacy threats to its training data [39], [13], [19], [12], [43].”
BadNL: Backdoor Attacks against NLP Models with Semantic-preserving Improvements (2020, background): “For instance, multiple works have explored the security and privacy threats of data used to train DNNmodels, such as membership inference attack [30, 35, 36], dataset reconstruction attack [33], and property inference attack [8, 12].”