Matthew Jagielski

发表

Subpopulation Data Poisoning Attacks pdf

Alina Oprea, Giorgio Severi, Matthew Jagielski, 2020, CCS.

Why Do Adversarial Attacks Transfer? Explaining Transferability of Evasion and Poisoning Attacks

Fabio Roli, Marco Melis, Cristina Nita-Rotaru, 2018, USENIX Security Symposium.

Differentially Private Fair Learning

Aaron Roth, Jonathan Ullman, Jieming Mao, 2018, ICML.

Extracting Training Data from Large Language Models pdf

Colin Raffel, Dawn Song, Eric Wallace, 2020, USENIX Security Symposium.

Subverting Fair Image Search with Generative Adversarial Perturbations

A. Ghosh, Matthew Jagielski, Chris L. Wilson, 2022, FAccT.

How to Combine Membership-Inference Attacks on Multiple Updated Models

Jonathan Ullman, Roxana Geambasu, Alina Oprea, 2022, ArXiv.

SafeNet: Mitigating Data Poisoning Attacks on Private Machine Learning

Alina Oprea, Matthew Jagielski, Harsh Chaudhari, 2022, IACR Cryptol. ePrint Arch..

Measuring Forgetting of Memorized Training Examples

Florian Tramèr, Nicolas Papernot, Nicholas Carlini, 2022, ArXiv.

SafeNet: The Unreasonable Effectiveness of Ensembles in Private Collaborative Learning

Alina Oprea, Matthew Jagielski, Harsh Chaudhari, 2022, 2023 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML).

Truth Serum: Poisoning Machine Learning Models to Reveal Their Secrets

Florian Tramèr, Nicholas Carlini, Hoang M. Le, 2022, CCS.

Network-Level Adversaries in Federated Learning

C. Nita-Rotaru, Alina Oprea, Matthew Jagielski, 2022, 2022 IEEE Conference on Communications and Network Security (CNS).

CIS 410 Final Report on Hidden Subgroup Problem

Matthew Jagielski, Zhimeng Wang, D. Roh, 2016 .

High Accuracy and High Fidelity Extraction of Neural Networks

Nicolas Papernot, Nicholas Carlini, David Berthelot, 2019, USENIX Security Symposium.

High-Fidelity Extraction of Neural Network Models pdf

David Berthelot, Nicholas Carlini, Nicolas Papernot, 2019, ArXiv.

Extracting Training Data from Large Language Models

Colin Raffel, Dawn Song, Eric Wallace, 2020, USENIX Security Symposium.

Poisoning Web-Scale Training Datasets is Practical

Christopher A. Choquette-Choo, Florian Tramèr, Nicholas Carlini, 2023, ArXiv.

Quantifying Memorization Across Neural Language Models

Florian Tramèr, Nicholas Carlini, Chiyuan Zhang, 2022, ICLR.

The Privacy Onion Effect: Memorization is Relative

Florian Tramèr, Nicolas Papernot, Nicholas Carlini, 2022, NeurIPS.

Challenges towards the Next Frontier in Privacy

Ryan M. Rogers, Weijie J. Su, Nicolas Papernot, 2023, ArXiv.

Auditing Differentially Private Machine Learning: How Private is Private SGD? pdf

Jonathan Ullman, Alina Oprea, Matthew Jagielski, 2020, NeurIPS.

On the Intriguing Connections of Regularization, Input Gradients and Transferability of Evasion and Poisoning Attacks pdf

Fabio Roli, Marco Melis, Cristina Nita-Rotaru, 2018, ArXiv.

Extracting Training Data from Diffusion Models

Florian Tramèr, Nicholas Carlini, Borja Balle, 2023, USENIX Security Symposium.

Students Parrot Their Teachers: Membership Inference on Model Distillation

Christopher A. Choquette-Choo, Nicholas Carlini, Katherine Lee, 2023, ArXiv.

Counterfactual Memorization in Neural Language Models pdf

Daphne Ippolito, Nicholas Carlini, Florian Tramer, 2021, ArXiv.

Manipulating Machine Learning: Poisoning Attacks and Countermeasures for Regression Learning

Chang Liu, Bo Li, Cristina Nita-Rotaru, 2018, 2018 IEEE Symposium on Security and Privacy (SP).

Measuring Forgetting of Memorized Training Examples

Florian Tramèr, Nicolas Papernot, Nicholas Carlini, 2022, ICLR.

Security of Machine Learning

Florian Tramèr, Nicholas Carlini, D. Song, 2019 .

SNAP: Efficient Extraction of Private Properties with Poisoning

Florian Tramèr, Jonathan Ullman, Alina Oprea, 2022, ArXiv.

PaLM 2 Technical Report

Andrew M. Dai, Jan A. Botha, Laurent El Shafey, 2023, ArXiv.

Debugging Differential Privacy: A Case Study for Privacy Auditing

Florian Tramèr, Nicholas Carlini, A. Terzis, 2022, ArXiv.

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

Aldo G. Carranza, Milad Nasr, N. Ponomareva, 2023, ArXiv.

Preventing Verbatim Memorization in Language Models Gives a False Sense of Privacy

Christopher A. Choquette-Choo, Florian Tramèr, Nicholas Carlini, 2022, ArXiv.

Privacy Auditing with One (1) Training Run

T. Steinke, Milad Nasr, Matthew Jagielski, 2023, ArXiv.

A Note On Interpreting Canary Exposure

Matthew Jagielski, 2023, ArXiv.

Tight Auditing of Differentially Private Machine Learning

Florian Tramèr, Nicholas Carlini, A. Terzis, 2023, USENIX Security Symposium.

Cryptanalytic Extraction of Neural Network Models

Ilya Mironov, Nicholas Carlini, Matthew Jagielski, 2020, CRYPTO.

Preventing Verbatim Memorization in Language Models Gives a False Sense of Privacy

Christopher A. Choquette-Choo, Florian Tramèr, Nicholas Carlini, 2022, ArXiv.

SNAP: Efficient Extraction of Private Properties with Poisoning

Florian Tramèr, Jonathan Ullman, Alina Oprea, 2022, 2023 IEEE Symposium on Security and Privacy (SP).

Q UANTIFYING M EMORIZATION A CROSS N EURAL L ANGUAGE M ODELS

Florian Tramèr, Nicholas Carlini, Katherine Lee, 2023 .

Threat Detection for Collaborative Adaptive Cruise Control in Connected Cars

Cristina Nita-Rotaru, Chung-Wei Lin, Shinichi Shiraishi, 2018, WISEC.

Are aligned neural networks adversarially aligned?

Pang Wei Koh, Christopher A. Choquette-Choo, Florian Tramèr, 2023, NeurIPS.

Randomness in ML Defenses Helps Persistent Attackers and Hinders Evaluators

Florian Tramèr, Nicholas Carlini, Lujo Bauer, 2023, ArXiv.

M EASURING F ORGETTING OF M EMORIZED T RAINING E XAMPLES

Florian Tramèr, Nicolas Papernot, Nicholas Carlini, 2023 .

PaLM 2 Technical Report

Andrew M. Dai, Jan A. Botha, Laurent El Shafey, 2023, ArXiv.

Extracting Training Data from Large Language Models

Tom B. Brown, Florian Tramèr, Nicholas Carlini, 2020, USENIX Security Symposium.

Privacy Side Channels in Machine Learning Systems

Christopher A. Choquette-Choo, Florian Tramèr, Nicholas Carlini, 2023, USENIX Security Symposium.

Counterfactual Memorization in Neural Language Models

Florian Tramèr, Nicholas Carlini, Katherine Lee, 2021, NeurIPS.

Preventing Generation of Verbatim Memorization in Language Models Gives a False Sense of Privacy

Florian Tramèr, Nicholas Carlini, Katherine Lee, 2022, INLG.

Backdoor Attacks for In-Context Learning with Language Models

Florian Tramèr, Nicholas Carlini, Matthew Jagielski, 2023, ArXiv.

This paper is included in the Proceedings of the 32nd USENIX Security

Princeton, Vikash Sehwag, Daphne Ippolito, 2023 .

PaLM 2 Technical Report

Andrew M. Dai, Jan A. Botha, Laurent El Shafey, 2023, ArXiv.

Secure Communication Channel Establishment: TLS 1.3 (over TCP Fast Open) vs. QUIC

Cristina Nita-Rotaru, Alexandra Boldyreva, Shan Chen, 2019, IACR Cryptol. ePrint Arch..

Synthetic Query Generation for Privacy-Preserving Deep Retrieval Systems using Differentially Private Language Models

Aldo G. Carranza, Milad Nasr, N. Ponomareva, 2023, NAACL.

Scalable Extraction of Training Data from (Production) Language Models

Christopher A. Choquette-Choo, Florian Tramèr, Nicholas Carlini, 2023, ArXiv.

Tight Auditing of Differentially Private Machine Learning

Florian Tramèr, Nicholas Carlini, A. Terzis, 2023, USENIX Security Symposium.