Picture for Matthew Jagielski

Matthew Jagielski

Beyond the Mean: Differentially Private Prototypes for Private Transfer Learning

Add code
Jun 12, 2024
Viaarxiv icon

Phantom: General Trigger Attacks on Retrieval Augmented Language Generation

Add code
May 30, 2024
Viaarxiv icon

Noise Masking Attacks and Defenses for Pretrained Speech Models

Apr 02, 2024
Viaarxiv icon

Scalable Extraction of Training Data from (Production) Language Models

Nov 28, 2023
Figure 1 for Scalable Extraction of Training Data from (Production) Language Models
Figure 2 for Scalable Extraction of Training Data from (Production) Language Models
Figure 3 for Scalable Extraction of Training Data from (Production) Language Models
Figure 4 for Scalable Extraction of Training Data from (Production) Language Models
Viaarxiv icon

Privacy Side Channels in Machine Learning Systems

Add code
Sep 11, 2023
Viaarxiv icon

Are aligned neural networks adversarially aligned?

Jun 26, 2023
Figure 1 for Are aligned neural networks adversarially aligned?
Figure 2 for Are aligned neural networks adversarially aligned?
Figure 3 for Are aligned neural networks adversarially aligned?
Figure 4 for Are aligned neural networks adversarially aligned?
Viaarxiv icon

A Note On Interpreting Canary Exposure

Jun 02, 2023
Viaarxiv icon

PaLM 2 Technical Report

Add code
May 17, 2023
Figure 1 for PaLM 2 Technical Report
Figure 2 for PaLM 2 Technical Report
Figure 3 for PaLM 2 Technical Report
Figure 4 for PaLM 2 Technical Report
Viaarxiv icon

Privacy Auditing with One (1) Training Run

Add code
May 15, 2023
Figure 1 for Privacy Auditing with One (1) Training Run
Figure 2 for Privacy Auditing with One (1) Training Run
Figure 3 for Privacy Auditing with One (1) Training Run
Figure 4 for Privacy Auditing with One (1) Training Run
Viaarxiv icon

Privacy-Preserving Recommender Systems with Synthetic Query Generation using Differentially Private Large Language Models

May 10, 2023
Figure 1 for Privacy-Preserving Recommender Systems with Synthetic Query Generation using Differentially Private Large Language Models
Figure 2 for Privacy-Preserving Recommender Systems with Synthetic Query Generation using Differentially Private Large Language Models
Figure 3 for Privacy-Preserving Recommender Systems with Synthetic Query Generation using Differentially Private Large Language Models
Figure 4 for Privacy-Preserving Recommender Systems with Synthetic Query Generation using Differentially Private Large Language Models
Viaarxiv icon