Apostolos Mavrogiannakis

PhD candidate at UC Santa Cruz. I build database systems that reveal nothing about the data they process.

Download CV See the research

What the cloud sees

Every run touches memory in the same fixed order. The pattern says nothing about the data.

Trusted hardware lets the cloud compute on data it cannot read. But the pattern of memory it touches still leaks. My research makes that pattern independent of the data, with oblivious algorithms and systems fast enough for real workloads: parallel, distributed, and now at terabyte scale.

Zero-leakage fully parallel joins.

USENIX Security 2025.

Oblivious parallel operators for foreign-key and non-foreign-key joins inside hardware enclaves. Nothing is revealed beyond input and output sizes, even to an adversary watching every memory access.

Up to 80×
faster than the sequential prior state of the art, running in parallel
Beyond 140GB
Oblivious joins demonstrated on datasets this large for the first time

Read about the paperePrintCode

Distributed oblivious joins, at terabyte scale.

SOSP 2026. Nominated for Best Paper Award.

The first scalable, distributed, and fully oblivious database system. DISCO combines hardware enclaves with oblivious primitives and optimizations that cut inter-server communication, unlocking parallel oblivious computation across machines.

DISCO: 1× the running time of DISCO; Jodes: 88× the running time of DISCO; SODA: 180× the running time of DISCO.

88×
faster than Jodes
180×
faster than SODA
Terabyte
scale reached by an oblivious system for the first time

Read about the paper

Publications

  1. DISCO: Distributed and Scalable Oblivious Joins and Oblivious Primitives

    Apostolos Mavrogiannakis, Xian Wang, Ioannis Demertzis, Dimitrios Papadopoulos, Minos Garofalakis

    ACM Symposium on Operating Systems Principles (SOSP), 2026. Nominated for Best Paper Award

    Abstract

  2. OBLIVIATOR: Oblivious Parallel Joins and Other Operators in Shared Memory Environments

    Apostolos Mavrogiannakis, Xian Wang, Ioannis Demertzis, Dimitrios Papadopoulos, Minos Garofalakis

    USENIX Security, 2025

    Abstract Paper ePrintCode

Research Experience

  • Graduate Student Researcher, UC Santa Cruz Santa Cruz, CA · September 2022 – Present

    Research on scalable oblivious databases and systems, advised by Prof. Ioannis Demertzis.

Talks & Presentations

  • October 2026

    Zero-Leakage Query Processing

    NorCal DB 2026

  • October 2026

    DISCO: Distributed and Scalable Oblivious Joins and Oblivious Primitives

    ACM SOSP, Prague, Czechia

  • September 2026

    DISCO: Distributed and Scalable Oblivious Joins and Oblivious Primitives

    FORTH, Heraklion, Greece

  • September 2025

    Scalable and Oblivious Database Systems

    Advancement to Candidacy, UC Santa Cruz

  • August 2025

    OBLIVIATOR: Oblivious Parallel Joins and other Operators in Shared Memory Environments

    USENIX Security

  • February 2025

    Oblivious Parallel Joins and other operators

    UC Santa Cruz Security Seminar

Professional Service

  • Journal reviewing

    TDSC (2024)

  • Conference external reviewer

    • USENIX Security Symposium (2024, 2025, 2026)
    • IEEE Symposium on Security and Privacy (2024, 2025, 2026)
    • IEEE European Symposium on Security and Privacy (2026)
    • ACM CCS (2023, 2024, 2026)
    • NDSS (2024)
    • SIGMOD (2024, 2025, 2026)
    • VLDB (2026)
    • ASIACCS (2023, 2024)
    • ICDE (2024, 2025)

News & Updates

  • August 2026

    Our paper DISCO got accepted in SOSP 2026 and is nominated for the Best Paper Award!

  • February 2025

    Our paper on oblivious databases got accepted in USENIX Security 2025!

  • September 2023

    Received a Research Grant from the Gerondelis Foundation.

Get in touch.

Feel free to reach out if you’re interested in my research or potential collaborations.

GitHubLinkedInGoogle Scholar