Data-Oblivious ML Accelerators Using Hardware Security Extensions

Hossam Elatali, John Z. Jekel, Lachlan J. Gunn, N. Asokan

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

Abstract

We present Dolma, which applies dynamic information flow tracking (DIFT) to the Gemmini matrix multiplication accelerator. With the BliMe CPU extensions, it efficiently guarantees client data confidentiality, even in the presence of malicious/vulnerable software and side channel attacks on servers.

Original languageEnglish
Title of host publicationProceedings of the 2024 IEEE International Symposium on Hardware Oriented Security and Trust, HOST 2024
PublisherIEEE
Pages373-377
Number of pages5
ISBN (Electronic)979-8-3503-7394-3
DOIs
Publication statusPublished - 2024
MoE publication typeA4 Conference publication
EventIEEE International Symposium on Hardware Oriented Security and Trust - McLean, United States
Duration: 6 May 20249 May 2024

Publication series

NameProceedings of the IEEE International Symposium on Hardware-Oriented Security and Trust
ISSN (Electronic)2765-8406

Conference

ConferenceIEEE International Symposium on Hardware Oriented Security and Trust
Abbreviated titleHOST
Country/TerritoryUnited States
CityMcLean
Period06/05/202409/05/2024

Fingerprint

Dive into the research topics of 'Data-Oblivious ML Accelerators Using Hardware Security Extensions'. Together they form a unique fingerprint.

Cite this