190 citations · 316 across the 24 of their papers we have counts for
29 papers
Distributed and Private Textual Data Synthesis from Embeddings
Ergute Bao, Hongyan Chang, Ali Shahin Shamsabadi +2
We revisit differentially private (DP) text synthesis in the realistic setting of distributed users, where privacy concerns preclude a trusted curator with access to raw user texts…
AgentStop: Terminating Local AI Agents Early to Save Energy in Consumer Devices
Dzung Pham, Kleomenis Katevas, Ali Shahin Shamsabadi +1
Autonomous agents powered by large language models (LLMs) are increasingly used to automate complex, multi-step tasks such as coding or web-based question answering. While remote,…
SPILLage: Agentic Oversharing on the Web
Jaechul Roh, Eugene Bagdasarian, Hamed Haddadi +1
LLM-powered agents are beginning to automate user's tasks across the open web, often with access to user resources such as emails and calendars. Unlike standard LLMs answering ques…
Frontier AI Auditing: Toward Rigorous Third-Party Assessment of Safety and Security Practices at Leading AI Companies
Miles Brundage, Noemi Dreksler, Aidan Homewood +45
We outline a vision for frontier AI auditing, which we define as rigorous third-party verification of frontier AI developers' safety and security claims, and evaluation of their sy…
Privacy Practices of Browser Agents
Alisha Ukani, Hamed Haddadi, Ali Shahin Shamsabadi +1
This paper presents a systematic evaluation of the privacy behaviors and attributes of eight recent, popular browser agents. Browser agents are software that automate Web browsing…
Membership and Memorization in LLM Knowledge Distillation
Ziqi Zhang, Ali Shahin Shamsabadi, Hanxiao Lu +2
Recent advances in Knowledge Distillation (KD) aim to mitigate the high computational demands of Large Language Models (LLMs) by transferring knowledge from a large ''teacher'' to…
Confidential Guardian: Cryptographically Prohibiting the Abuse of Model Abstention
Stephan Rabanser, Ali Shahin Shamsabadi, Olive Franzese +3
Cautious predictions -- where a machine learning model abstains when uncertain -- are crucial for limiting harmful errors in safety-critical applications. In this work, we identify…
NoEsis: Differentially Private Knowledge Transfer in Modular LLM Adaptation
Rob Romijnders, Stefanos Laskaridis, Ali Shahin Shamsabadi +1
Large Language Models (LLM) are typically trained on vast amounts of data from various sources. Even when designed modularly (e.g., Mixture-of-Experts), LLMs can leak privacy on th…
Secure and Confidential Certificates of Online Fairness
Olive Franzese, Ali Shahin Shamsabadi, Carter Luck +1
The black-box service model enables ML service providers to serve clients while keeping their intellectual property and client data confidential. Confidentiality is critical for de…
Context-Aware Membership Inference Attacks against Pre-trained Large Language Models
Hongyan Chang, Ali Shahin Shamsabadi, Kleomenis Katevas +2
Membership Inference Attacks (MIAs) on pre-trained Large Language Models (LLMs) aim at determining if a data point was part of the model's training set. Prior MIAs that are built f…
Identifying and Mitigating Privacy Risks Stemming from Language Models: A Survey
Victoria Smith, Ali Shahin Shamsabadi, Carolyn Ashurst +1
Large Language Models (LLMs) have shown greatly enhanced performance in recent years, attributed to increased size and extensive training data. This advancement has led to widespre…
Reconstructing Individual Data Points in Federated Learning Hardened with Differential Privacy and Secure Aggregation
Franziska Boenisch, Adam Dziedzic, Roei Schuster +3
Federated learning (FL) is a framework for users to jointly train a machine learning model. FL is promoted as a privacy-enhancing technology (PET) that provides data minimization:…
Private Multi-Winner Voting for Machine Learning
Adam Dziedzic, Christopher A Choquette-Choo, Natalie Dullerud +6
Private multi-winner voting is the task of revealing -hot binary vectors satisfying a bounded differential privacy (DP) guarantee. This task has been understudied in machine lea…
On the reversibility of adversarial attacks
Chau Yi Li, Ricardo Sánchez-Matilla, Ali Shahin Shamsabadi +2
Adversarial attacks modify images with perturbations that change the prediction of classifiers. These modified images, known as adversarial examples, expose the vulnerabilities of…
GAP: Differentially Private Graph Neural Networks with Aggregation Perturbation
Sina Sajadmanesh, Ali Shahin Shamsabadi, Aurélien Bellet +1
In this paper, we study the problem of learning Graph Neural Networks (GNNs) with Differential Privacy (DP). We propose a novel differentially private GNN based on Aggregation Pert…
Differentially Private Speaker Anonymization
Ali Shahin Shamsabadi, Brij Mohan Lal Srivastava, Aurélien Bellet +5
Sharing real-world speech utterances is key to the training and deployment of voice-based services. However, it also raises privacy risks as speech contains a wealth of personal da…
Tubes Among Us: Analog Attack on Automatic Speaker Identification
Shimaa Ahmed, Yash Wani, Ali Shahin Shamsabadi +4
Recent years have seen a surge in the popularity of acoustics-enabled personal devices powered by machine learning. Yet, machine learning has proven to be vulnerable to adversarial…
When the Curious Abandon Honesty: Federated Learning Is Not Private
Franziska Boenisch, Adam Dziedzic, Roei Schuster +3
In federated learning (FL), data does not leave personal devices when they are jointly training a machine learning model. Instead, these devices share gradients, parameters, or oth…
FoolHD: Fooling speaker identification by Highly imperceptible adversarial Disturbances
Ali Shahin Shamsabadi, Francisco Sepúlveda Teixeira, Alberto Abad +3
Speaker identification models are vulnerable to carefully designed adversarial perturbations of their input signals that induce misclassification. In this work, we propose a white-…
Exploiting vulnerabilities of deep neural networks for privacy protection
Ricardo Sanchez-Matilla, Chau Yi Li, Ali Shahin Shamsabadi +2
Adversarial perturbations can be added to images to protect their content from unwanted inferences. These perturbations may, however, be ineffective against classifiers that were n…
DarkneTZ: Towards Model Privacy at the Edge using Trusted Execution Environments
Fan Mo, Ali Shahin Shamsabadi, Kleomenis Katevas +4
We present DarkneTZ, a framework that uses an edge device's Trusted Execution Environment (TEE) in conjunction with model partitioning to limit the attack surface against Deep Neur…
PrivEdge: From Local to Distributed Private Training and Prediction
Ali Shahin Shamsabadi, Adria Gascon, Hamed Haddadi +1
Machine Learning as a Service (MLaaS) operators provide model training and prediction on the cloud. MLaaS applications often rely on centralised collection and aggregation of user…
ColorFool: Semantic Adversarial Colorization
Ali Shahin Shamsabadi, Ricardo Sanchez-Matilla, Andrea Cavallaro
Adversarial attacks that generate small L_p-norm perturbations to mislead classifiers have limited success in black-box settings and with unseen classifiers. These attacks are also…
EdgeFool: An Adversarial Image Enhancement Filter
Ali Shahin Shamsabadi, Changjae Oh, Andrea Cavallaro
Adversarial examples are intentionally perturbed images that mislead classifiers. These images can, however, be easily detected using denoising algorithms, when high-frequency spat…