MPI-INF Logo
Campus Event Calendar

Event Entry

What and Who

Oblivious Vector Databases

Sepehr Azardar
University of Teheran
PhD Application Talk
AG 1, AG 2, AG 3, INET, AG 4, AG 5, D6, SWS, RG1, MMCI  
AG Audience
English

Date, Time and Location

Tuesday, 28 January 2025
13:30
30 Minutes
Virtual talk
zoom

Abstract

With the increasing adoption of vector databases in various machine learning and data-driven applications, ensuring privacy and security while maintaining efficiency has become a critical challenge. A key concern is the exposure of access patterns, which can inadvertently reveal sensitive information about queries and underlying data. In this talk, I will present my work on securing vector databases against access pattern attacks by integrating obfuscation techniques into indexing mechanisms. Specifically, I focus on applying these techniques to hierarchical navigable small-world (HNSW) graphs, a state-of-the-art indexing method. My approach strikes a balance between privacy and performance, offering a scalable solution for practical applications. Through this research, I aim to demonstrate how combining advanced obfuscation strategies with robust indexing methods can mitigate privacy risks without compromising the system's efficiency.

Contact

Ina Geisler
+49 681 9325 1802
--email hidden

Virtual Meeting Details

Zoom
passcode not visible
logged in users only

Ina Geisler, 01/24/2025 11:37 -- Created document.