Project Repository for ECE209AS (Winter 2020) Course Project by Swapnil Sayan Saha, Brian Wang and Vivek Jain. Supervisor: Dr. Mani Srivastava, Professor of ECE and CS, UCLA
The goal of Project P2I (Privacy Preserving Inferencing) is to benchmark and implement state-of-the-art secure multiparty computation (SMPC) and aggregation protocols in multi-layer medical cyberphysical systems (MCPS) void of centralized cloud inferencing and enabling secure inferencing at the edge. Project P2I forms the basis and starting-point for enabling privacy preserving inferencing at the edge for collaborative computing in MCPS, using existing state-of-the-art secure aggregation and SMPC protocols without the notion of TTP.
- Our benchmark shows that it is possible to solve classical MPC (and SA) problems and queries on resource-constrained edge devices using existing state of the art MPC (and SA) protocols, without sacrificing security gurantees.
- Our proposed architectures do not require any external party in the pipeline, yet achieving collaborative computing goals and comparable to all existing SMPC or aggregation architectures proposed in literature.
- A wide variety of custom oblivious functions can be designed to suit one's needs.
- Our code is easily integrable into existing systems as a black box in the middle.
- Our chosen MPC and SA protocols are secure from an information theoretic-setting, with an additional layer of AES on top and communication being done via secure protocols like SSL, hence ensuring security for all layers of a MCPS.
Project Website with Results: https://bvs209as.weebly.com/