While modern machine learning models rely on increasingly large training datasets, data is
often limited in privacy-sensitive domains. Generative models trained with differential privacy …

Privacy concerns have attracted increasing attention in data-driven products due to the
tendency of machine learning models to memorize sensitive training data. Generating …

Differentially private (DP) optimization is the standard paradigm to learn large neural
networks that are accurate and privacy-preserving. The computational cost for DP deep …

Recent trends in the modus operandi of technologically-aware criminal groups engaged in
illicit goods trafficking (eg, firearms, drugs, cultural artifacts, etc.) have given rise to …

We study the problem of differentially private (DP) fine-tuning of large pre-trained models–a
recent privacy-preserving approach suitable for solving downstream tasks with sensitive …

Differentially private (DP) training preserves the data privacy usually at the cost of slower
convergence (and thus lower accuracy), as well as more severe mis-calibration than its non …

Artificial intelligence (AI) has seen a tremendous surge in capabilities thanks to the use of
foundation models trained on internet-scale data. On the flip side, the uncurated nature of …

Data augmentation techniques, such as image transformations and combinations, are highly
effective at improving the generalization of computer vision models, especially when training …

Visual Prompting (VP) is an emerging and powerful technique that allows sample-efficient
adaptation to downstream tasks by engineering a well-trained frozen source model. In this …

Differentially private stochastic gradient descent (DP-SGD) is the workhorse algorithm for
recent advances in private deep learning. It provides a single privacy guarantee to all …