Alert button
Picture for Eugene Bagdasaryan

Eugene Bagdasaryan

Alert button

Ceci n'est pas une pomme: Adversarial Illusions in Multi-Modal Embeddings

Aug 22, 2023
Eugene Bagdasaryan, Vitaly Shmatikov

Viaarxiv icon

(Ab)using Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs

Jul 24, 2023
Eugene Bagdasaryan, Tsung-Yin Hsieh, Ben Nassi, Vitaly Shmatikov

Figure 1 for (Ab)using Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs
Figure 2 for (Ab)using Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs
Figure 3 for (Ab)using Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs
Figure 4 for (Ab)using Images and Sounds for Indirect Instruction Injection in Multi-Modal LLMs
Viaarxiv icon

Hyperparameter Search Is All You Need For Training-Agnostic Backdoor Robustness

Feb 09, 2023
Eugene Bagdasaryan, Vitaly Shmatikov

Figure 1 for Hyperparameter Search Is All You Need For Training-Agnostic Backdoor Robustness
Figure 2 for Hyperparameter Search Is All You Need For Training-Agnostic Backdoor Robustness
Figure 3 for Hyperparameter Search Is All You Need For Training-Agnostic Backdoor Robustness
Figure 4 for Hyperparameter Search Is All You Need For Training-Agnostic Backdoor Robustness
Viaarxiv icon

Training a Tokenizer for Free with Private Federated Learning

Mar 15, 2022
Eugene Bagdasaryan, Congzheng Song, Rogier van Dalen, Matt Seigel, Áine Cahill

Figure 1 for Training a Tokenizer for Free with Private Federated Learning
Figure 2 for Training a Tokenizer for Free with Private Federated Learning
Figure 3 for Training a Tokenizer for Free with Private Federated Learning
Figure 4 for Training a Tokenizer for Free with Private Federated Learning
Viaarxiv icon

Spinning Language Models for Propaganda-As-A-Service

Dec 09, 2021
Eugene Bagdasaryan, Vitaly Shmatikov

Figure 1 for Spinning Language Models for Propaganda-As-A-Service
Figure 2 for Spinning Language Models for Propaganda-As-A-Service
Figure 3 for Spinning Language Models for Propaganda-As-A-Service
Figure 4 for Spinning Language Models for Propaganda-As-A-Service
Viaarxiv icon

Towards Sparse Federated Analytics: Location Heatmaps under Distributed Differential Privacy with Secure Aggregation

Nov 03, 2021
Eugene Bagdasaryan, Peter Kairouz, Stefan Mellem, Adrià Gascón, Kallista Bonawitz, Deborah Estrin, Marco Gruteser

Figure 1 for Towards Sparse Federated Analytics: Location Heatmaps under Distributed Differential Privacy with Secure Aggregation
Figure 2 for Towards Sparse Federated Analytics: Location Heatmaps under Distributed Differential Privacy with Secure Aggregation
Figure 3 for Towards Sparse Federated Analytics: Location Heatmaps under Distributed Differential Privacy with Secure Aggregation
Figure 4 for Towards Sparse Federated Analytics: Location Heatmaps under Distributed Differential Privacy with Secure Aggregation
Viaarxiv icon

Spinning Sequence-to-Sequence Models with Meta-Backdoors

Jul 22, 2021
Eugene Bagdasaryan, Vitaly Shmatikov

Figure 1 for Spinning Sequence-to-Sequence Models with Meta-Backdoors
Figure 2 for Spinning Sequence-to-Sequence Models with Meta-Backdoors
Figure 3 for Spinning Sequence-to-Sequence Models with Meta-Backdoors
Viaarxiv icon

Blind Backdoors in Deep Learning Models

May 08, 2020
Eugene Bagdasaryan, Vitaly Shmatikov

Figure 1 for Blind Backdoors in Deep Learning Models
Figure 2 for Blind Backdoors in Deep Learning Models
Figure 3 for Blind Backdoors in Deep Learning Models
Figure 4 for Blind Backdoors in Deep Learning Models
Viaarxiv icon

Decentralized Policy-Based Private Analytics

Mar 14, 2020
Kleomenis Katevas, Eugene Bagdasaryan, Jason Waterman, Mohamad Mounir Safadieh, Hamed Haddadi, Deborah Estrin

Figure 1 for Decentralized Policy-Based Private Analytics
Figure 2 for Decentralized Policy-Based Private Analytics
Figure 3 for Decentralized Policy-Based Private Analytics
Figure 4 for Decentralized Policy-Based Private Analytics
Viaarxiv icon

Salvaging Federated Learning by Local Adaptation

Feb 12, 2020
Tao Yu, Eugene Bagdasaryan, Vitaly Shmatikov

Figure 1 for Salvaging Federated Learning by Local Adaptation
Figure 2 for Salvaging Federated Learning by Local Adaptation
Figure 3 for Salvaging Federated Learning by Local Adaptation
Figure 4 for Salvaging Federated Learning by Local Adaptation
Viaarxiv icon