Privacy

One-shot label-only attack (OSLO) achieves high membership inference accuracy with only one query, surpassing existing methods by a large margin.

This paper introduces computationally efficient algorithms for differentially private learning by leveraging public data, overcoming previous computational limitations and enabling broader practical a…

This paper proposes a computationally efficient, differentially private best subset selection method for high-dimensional sparse linear regression, achieving both strong statistical utility and provab…

Public data boosts private AI accuracy even with extreme distribution shifts, improving private model training by up to 67% in three tasks. This is due to shared low-dimensional representations betwe…

Learnware systems enable model reuse; this paper proves RKME specifications protect developers’ training data while enabling effective model identification.

Graph representation learning’s structural vulnerabilities are proven and mitigated via noisy aggregation, revealing crucial privacy-utility trade-offs.

New algorithms achieve near-optimal differentially private U-statistic estimation, significantly improving accuracy over existing methods.

This paper presents novel measures quantifying data easiness for DP subspace estimation, supporting them with improved upper and lower bounds and a practical algorithm.

Near-optimal algorithm for private resource allocation is introduced, achieving improved accuracy and privacy guarantees.

Meta-learning and differential privacy combine to enable accurate, well-calibrated private regression, even with limited data, via the novel DPConvCNP model.

Nimbus achieves 2.7-4.7x speedup in BERT base inference using novel two-party computation techniques for efficient matrix multiplication and non-linear layer approximation.

This paper presents a new auditing method for DP-SGD that provides substantially tighter black-box privacy analyses than previous methods, yielding significantly closer empirical estimates to theoreti…

This research unveils a fundamental interplay between local differential privacy (LDP) and robustness against data corruption and heavy-tailed rewards in multi-armed bandits, offering a tight characte…

Langevin unlearning offers a novel, privacy-preserving machine unlearning framework based on noisy gradient descent, handling both convex and non-convex problems efficiently.

New algorithm improves differentially private estimations by adapting to dataset hardness, enhancing accuracy for variance, classification, and regression tasks.

Instance-optimal private density estimation algorithms, adapting to data characteristics for improved accuracy in the Wasserstein distance, are introduced.

Faster differentially private top-k selection achieved via a novel joint exponential mechanism with pruning, reducing time complexity from O(dk) to O(d+k²/ɛlnd).

Faster algorithms achieve optimal excess risk in user-level private stochastic convex optimization, overcoming limitations of prior methods without restrictive assumptions.

Researchers reveal a high risk of training data extraction from molecular pre-trained models, challenging the assumption that model sharing alone adequately protects against data theft.

This paper provides the first exact minimax-optimal mechanisms for locally differentially private sampling, applicable across all f-divergences.