Selected Publications

We introduce two protocols to calculate the contributivity to an aggregated model of the participants in a Federated Learning network. This project lays the foundations to leverage contributivity in FL over decentralised appplications.
AIChain 2020

A benchmark of recent privacy preserving ML solutions for secure inference. We compare frameworks powered by Homomorphic Encryption, Secure MPC and Intel SGX over 3 computer vision models and 2 datasets: MNIST and Malaria cells
PPMLP’20 @ CCS

A new general framework for privacy preserving deep learning. Focuses on ownership and secure processing of data and introduces a valuable representation based on chains of commands and tensors. The abstraction is used implement complex privacy preserving constructs such as Federated Learning, Secure Multiparty Computation, and Differential Privacy with a familiar deep learning API.
PPML @NeurIPS

Recent Publications

(A full list is available here: https://orcid.org/0000-0003-3178-9502

  • 2CP: Decentralized Protocols to Transparently Evaluate Contributivity in Blockchain Federated Learning Environments

    Details PDF

  • A Systematic Comparison of Encrypted Machine Learning Solutions for Image Classification

    Details PDF

  • A blockchain-orchestrated Federated Learning architecture for healthcare consortia

    Details PDF

  • A generic framework for privacy preserving deep learning

    Details PDF Code

  • Reproducible large-scale neuroimaging studies with the OpenMOLE workflow management system

    Details PDF Code Dataset Project

Recent & Upcoming Talks

Projects

Donaco

Donaco is an Imperial College London based startup that aims to facilitate online donations using machine learning.

iEx.ec DApp

OpenMOLE scientific platform, decentralised model exploration and parameter tuning on the iEx.ec network

OpenMOLE

Exploration, calibration and diagnosis of numerical model leveraging distributed computing environments.

GridScale

Scala library for accessing various file, batch systems, job schedulers and grid middlewares.

Teaching

Imperial College London (UK)

PhD student supervion

  • Dmitrii Usynin (2020-) - co-supervised with Ben Glocker (ICL) and Georgios Kaissis (TU Munich)

Labs and tutorials

  • Computer Graphics: OpenGL
  • Functional programming in Haskell: Introduction to functional programming
  • Java: Introduction to the language

Student projects supervised

ISIMA (Engineering School, France)

Courses led

  • Software Engineering 101: Git, Debugging with gdb
  • C++: Unit Testing (gtest), Template Meta-Programming, Standard Library
  • UML: Design Pattern, Class, State, Sequence Diagrams
  • Java: Java 7, Swing
  • Distributed computing: Cluster (PBS), Grid (EGI)
  • GPU Computing: CUDA, SIMD architecture

Student projects supervised

  • Various MEng projects supervised covering GPU parallelisation
  • Some associated publications (1, 2)

MSc in Computer Science (Blaise Pascal University, France)

Courses led

  • Introduction to parallel programming with CUDA

Contact