Differentially private algorithms protect model explanations from leaking training data.
problem Model explanations can leak training data, compromising privacy.
method Adaptive differentially private gradient descent algorithm to produce accurate, private explanations.
result Privacy amplification and reduction of overall privacy loss on explanation data.
DP-SGD analysis shows many datapoints leak less privacy than previously thought.
problem Empirical evidence suggests DP-SGD leaks less privacy than current analysis predicts.
method Developed a per-instance DP analysis of DP-SGD, introducing dependence on dataset distribution.
result Formally shows DP-SGD leaks significantly less privacy for many datapoints on common benchmarks.
New analysis shows SGD with noise doesn't leak more privacy with more iterations.
problem Privacy loss in noisy SGD with more iterations.
method Privacy Amplification by Iteration and Sampled Gaussian Mechanism.
result Privacy loss remains constant after a burn-in period, not increasing with more iterations.
Privacy and transparency are two key foundations of trustworthy machine learning. Model explanations offer insights into a model's decisions on input data, whereas privacy is primarily concerned with protecting information about the training data. We analyze connections between model explanations and the leakage of sen…
Active learning holds promise of significantly reducing data annotation costs while maintaining reasonable model performance. However, it requires sending data to annotators for labeling. This presents a possible privacy leak when the training set includes sensitive user data. In this paper, we describe an approach for…
Multi-task learning (MTL) refers to the paradigm of learning multiple related tasks together. In contrast, in single-task learning (STL) each individual task is learned independently. MTL often leads to better trained models because they can leverage the commonalities among related tasks. However, because MTL algorithm…
This paper studies trade-offs in private prediction methods.
problem Leakage of training data information in machine learning predictions.
method Private training and private prediction methods with trade-offs.
result Private training methods outperform private prediction methods in various settings.
Develops DP-SCD for stochastic coordinate descent, making it differentially private.
problem Privacy leak in auxiliary information during stochastic coordinate descent training.
method Develops DP-SCD, leveraging independent noise addition and decoupling/parallelizing coordinate updates.
result Demonstrates competitive performance against DP-SGD with less tuning.
Machine learning has been widely applied to various applications, some of which involve training with privacy-sensitive data. A modest number of data breaches have been studied, including credit card information in natural language data and identities from face dataset. However, most of these studies focus on supervise…
Federated learning leaks participant dataset quality even with secure aggregation.
problem Leakage of participant dataset quality in federated learning with secure aggregation.
method Image recognition experiments to infer and attribute dataset quality.
result Relative quality ordering of participants can be inferred and used for various purposes.
DarkneTZ protects edge devices from DNN model leaks using TEE and model partitioning.
problem Privacy risks of pre-trained DNNs on edge devices through membership inference attacks.
method Model partitioning into sensitive and untrusted parts, leveraging TEE.
result DarkneTZ provides reliable model privacy with minimal performance overhead.
Proposes QNN to protect input privacy in neural networks.
problem Protecting input privacy in neural networks.
method Quaternion-valued neural network (QNN) to hide input information.
result QNN effectively protects input privacy without significant accuracy loss.
Paper evaluates and improves private feature selection methods.
problem Feature selection in high-dimensional datasets with privacy constraints.
method Correlations-based order statistic privatized for feature selection.
result Our method significantly outperforms established baseline for private feature selection.
Proposes a privacy-preserving recommendation system using matrix factorization and differential privacy.
problem Privacy leakage in recommendation systems when anonymizing user data is not sufficient.
method Uses matrix factorization and differential privacy via the Gaussian mechanism.
result Demonstrates excellent utility for privacy-preserving recommendation systems.
For graphs generated from stochastic blockmodels, adjacency spectral embedding is asymptotically consistent. Further, adjacency spectral embedding composed with universally consistent classifiers is universally consistent to achieve the Bayes error. However when the graph contains private or sensitive information, trea…
Latent Dirichlet Allocation (LDA) is a popular topic modeling technique for discovery of hidden semantic architecture of text datasets, and plays a fundamental role in many machine learning applications. However, like many other machine learning algorithms, the process of training a LDA model may leak the sensitive inf…
Proposes element-level differential privacy for better privacy and utility in statistical learning.
problem Challenges of strong differential privacy in statistical learning applications.
method Introduces element-level differential privacy, extending classical DP to protect specific user elements.
result Provides better utility and more robust results compared to classical DP by allowing finer privacy protections.
Synthetic data can amplify privacy in linear regression models.
problem Understanding how synthetic data can enhance privacy in linear regression models.
method Investigated through the linear regression framework, analyzing synthetic data generated from random inputs and controlled inputs.
result Releasing a limited number of synthetic data points amplifies privacy beyond the model's inherent guarantees when inputs are random, but not when inputs are controlled by an adversary.
Machine learning on encrypted data has received a lot of attention thanks to recent breakthroughs in homomorphic encryption and secure multi-party computation. It allows outsourcing computation to untrusted servers without sacrificing privacy of sensitive data. We propose a practical framework to perform partially encr…
Our work investigates how to identify privacy violations in models using finite adversaries.
problem Identifying privacy violations in models with limited adversary capabilities.
method Investigates requirements for finite adversaries to identify privacy violations.
result Parameters quantify the capabilities of finite adversaries.
New approach protects privacy of deleted records in machine learning.
problem Privacy of deleted records in machine learning models.
method Sound deletion guarantee and noisy gradient descent algorithm.
result Privacy of existing records is necessary for deleted records' privacy.
Language models learn from training data and can leak private information.
problem Language models lack context understanding and can expose private data.
method Discussing the limitations of current privacy protection methods for language models.
result Existing privacy protection methods are insufficient for language models.
New algorithms protect user-level privacy in learning tasks.
problem Protecting user-level privacy in learning tasks with stringent constraints.
method Proposes algorithms for learning tasks under user-level differential privacy constraints.
result Privacy cost decreases as users provide more samples or as the number of users increases.
Many reinforcement learning applications involve the use of data that is sensitive, such as medical records of patients or financial information. However, most current reinforcement learning methods can leak information contained within the (possibly sensitive) data on which they are trained. To address this problem, w…
Differential privacy is a strong notion for privacy that can be used to prove formal guarantees, in terms of a privacy budget, ε, about how much information is leaked by a mechanism. However, implementations of privacy-preserving machine learning often select large values of ε in order to get acceptable utility of …
The paper investigates how data imbalance affects fairness and accuracy in differentially private deep learning.
problem Impact of data imbalance on fairness and accuracy in differentially private deep learning.
method Study the effects of different levels of imbalance in the data on the accuracy and fairness of decisions made by a model trained with differential privacy.
result Small imbalances and loose privacy guarantees can cause disparate impacts on model accuracy and fairness.
TVineSynth generates synthetic data to balance privacy and utility.
problem Balancing privacy and utility in synthetic data generation.
method Uses vine copula with truncation to control privacy and utility trade-off.
result Achieves superior privacy-utility balance compared to competitors.
Foundation models leak sensitive data in synthetic tabular data generation, especially LLaMA 3.3 70B.
problem Privacy leakage in synthetic tabular data generation using foundation models.
method Benchmarked three foundation models (GPT-4o-mini, LLaMA 3.3 70B, TabPFN v2) against four baselines on 35 real-world tables.
result Foundation models, especially LLaMA 3.3 70B, have the highest privacy risk in synthetic tabular data generation.
FEDMD-NFDP improves federated learning privacy without sacrificing performance.
problem Privacy leakage in federated learning when sharing predictions.
method Noise-Free Differential Privacy (NFDP) applied to federated model distillation.
result FEDMD-NFDP achieves comparable utility and privacy guarantees.
Survey on privacy issues in deep learning and proposed solutions.
problem Privacy concerns in deep learning models due to sensitive data.
method Review of existing privacy techniques and gaps in research.
result Identification of test-time inference privacy as a research gap.
Paper reconstructs training data from a single gradient query.
problem Privacy threats in federated learning due to model gradients.
method Provable attack using tensor decomposition.
result Training samples can be fully reconstructed from a single gradient query.
Researchers found PP-GANs can hide sensitive data in sanitized images, undermining privacy checks.
problem Lack of formal proofs of privacy in PP-GANs for image sanitization.
method Subverted PP-GANs for facial expression recognition to hide sensitive data in sanitized images.
result It is possible to hide sensitive identification data in sanitized PP-GAN output images, even allowing reconstruction of entire input images.
Paper shows Cox model optimisation leaks patient data in distributed learning.
problem Risk of patient data leakage in distributed learning models.
method Optimisation of Cox survival model using federated learning.
result Optimisation process can leak patient data, necessitating new validation methods.
This research examines rare spurious correlations in neural networks and their impact on accuracy and privacy.
problem Rare spurious correlations in neural networks and their privacy risks.
method Introducing spurious patterns correlated with a fixed class to a few training examples, analyzing ℓ2 regularization and Gaussian noise. result Rare spurious correlations can significantly impact neural network accuracy and privacy, and specific mitigation methods can be effective.
This paper improves deep learning models' accuracy with differential privacy using gradient encoding and denoising.
problem Deep learning models leak sensitive information about their training datasets.
method Gradient encoding to map gradients to a smaller vector space, and denoising for post-processing.
result Our technique achieves better model accuracy with differential privacy guarantees compared to state-of-the-art methods.
RENNs protect input privacy by rotating d-ary features.
problem Protecting input privacy from intermediate-layer features.
method Rotation-equivariant neural networks using d-ary vectors/tensors.
result RENNs effectively hide input information without degrading output accuracy.
Secure XGB for privacy-preserving machine learning in federated learning.
problem Privacy-preserving machine learning in federated learning with practical gradient tree boosting models.
method Secure multi-party computation, distributed model storage, secure permutation protocols.
result Our XGB models provide competitive accuracy and practical performance.
PRIME algorithm estimates mean while ensuring privacy and robustness.
problem Privacy and robustness in shared data analysis.
method Introduces PRIME, the first efficient algorithm for both privacy and robustness.
result Achieves both privacy and robustness for a wide range of distributions.
Collaborative learning allows participants to jointly train a model without data sharing. To update the model parameters, the central server broadcasts model parameters to the clients, and the clients send updating directions such as gradients to the server. While data do not leave a client device, the communicated gra…
This paper analyzes FL privacy risks and defensive strategies.
problem Privacy leakage in federated learning.
method Literature review of attack methods and defensive strategies.
result No single defensive strategy is sufficient for all attacks.
Analysis of model updates reveals sensitive data leaks.
problem Information leakage during model updates.
method Differential analysis of language model snapshots.
result New metrics (differential score, differential rank) reveal sensitive data leaks.
Enhances privacy in multimodal emotion recognition models.
problem Privacy leakage in multimodal emotion recognition models.
method Adversarial learning to unlearn private information.
result Improved privacy without significant performance loss.
Differential privacy for simple linear regression protects small datasets from individual data leaks.
problem Protecting sensitive personal information in small datasets from individual data leaks.
method Differential privacy algorithms for simple linear regression tailored for small datasets (tens to hundreds of datapoints).
result Robust estimators like Theil-Sen perform well on small datasets, but standard algorithms improve as dataset size increases.
Models leak information about their training data. This enables attackers to infer sensitive information about their training sets, notably determine if a data sample was part of the model's training set. The existing works empirically show the possibility of these membership inference (tracing) attacks against complex…
Paper introduces differential pairwise privacy for secure metric learning.
problem Securely measuring similarities of individuals given sensitive pairwise data.
method Develops differential pairwise privacy (DPP) to protect sensitive pairwise data.
result Achieves pairwise data privacy without significant performance loss.
Motivation: Human genomic datasets often contain sensitive information that limits use and sharing of the data. In particular, simple anonymisation strategies fail to provide sufficient level of protection for genomic data, because the data are inherently identifiable. Differentially private machine learning can help b…
HDP-VFL hybridizes DP for VFL, reducing privacy costs.
problem Privacy-preserving collaborative learning from vertically partitioned data.
method Hybrid DP framework combining HE and MPC for VFL.
result Achieves DP and JDP with negligible training time and accuracy trade-offs.
Strategic information is valuable either by remaining private (for instance if it is sensitive) or, on the other hand, by being used publicly to increase some utility. These two objectives are antagonistic and leaking this information might be more rewarding than concealing it. Unlike classical solutions that focus on …