Ramy E. Ali
 |
|
I am a Staff Researcher in the Multi-media and Cellular lab at Samsung, San Diego working on next generation wireless communication and machine learning systems. I was a postdoctoral researcher at the Electrical Engineering Department, University of Southern California working with Prof. Salman Avestimehr from 2020 to 2022. I currently work on secure and private machine learning, model compression and wireless communications.
News and Talks
February ’24: All Rivers Run to the Sea: Private Learning with Asymmetric Flows has been accepted at CVPR and also presented at ITA Slides
November ’22: Securing Secure Aggregation: Mitigating Multi-Round Privacy Leakage in Federated Learning has been accepted at AAAI-2023.
May ’22: Our work "3LegRace: Privacy-Preserving DNN Training over TEEs and GPUs” has been accepted at PETS’22.
March ’22: Our work, InfoCommit: Information-Theoretic Polynomial Commitment, has been accepted at T-IFS.
January ’22: Our work, LightSecAgg: a Lightweight and Versatile Design for Secure Aggregation in Federated Learning, has been accepted at MLSys 2022.
January ’22: Securing Secure Aggregation: Mitigating Multi-Round Privacy Leakage in Federated Learning has been accepted at FL-AAAI workshop, 2022.
December ’21: Verifiable Coded Computing: Towards Fast, Secure and Private Distributed Machine Learning has been accepted at IPDPS 2022.
December ’21: I've presented Multi-RoundSecAgg, at the Private AI Institute (Slides)
November ’21: ApproxIFER: A Model-Agnostic Approach to Resilient and Robust Prediction Serving Systems has been accepted at AAAI-2022.
November ’21: I've presented LightSecAgg, along with Jinhyun So, Chaoyang He and Salman Avestimehr at Facebook AI Research (FAIR) (Slides)
October ’21: Our work on asynchronous secure aggregation, Secure Aggregation for Buffered Asynchronous Federated Learning, has been accepted at NeurIPS NFFL Workshop.
October ’21: A preprint on leveraging TEEs in privacy-preserving ML, AsymML: An Asymmetric Decomposition Framework for Privacy-Preserving DNN Training and Inference
July ’21: List-Decodable Coded Computing: Breaking the Adversarial Toleration Barrier has been accepted in JSAIT.
January ’21: Consistency Analysis of Replication-Based Probabilistic Key-Value Stores has been accepted in ICC 2021.
October ’20: Our work, On Polynomial Approximations for Privacy-Preserving and Verifiable ReLU Networks has been accepted in the NeurIPS Workshop on Privacy Preserving Machine Learning (PPML & PRIML) (Extended Version, Slides).