Leveraging trusted execution environments for privacy aware mobility prediction
Mobile phones contains a hardware-based Trusted execution Environment (TEE), however, there use has been primarily limited to tasks such as secure boot. Until recently, the standardisation of the TEE interfaces and development and release of Open Source TEE (OP-TEE) has provided developers access to these trusted zones. The goal of this project will be to leverage the TEE for designing a system for predicting mobility of users in a privacy aware manner. The TEE makes it possible to keep the application provided by service providers opaque to the users and user data is opaque to the service providers. Thereby preserving the privacy of both the parties involved.
This phase will consist of designing the overall architecture of the system, allocating tasks to the trusted and the non trusted part of the operating system. You will deploy a basic prediction application based on logistic regression in the trusted zone. The end goal will be to deploy a fully working machine learning model in the trusted zone which will be trained on the incoming location data stream of a user.
You will work with trusted zone present in Raspberry Pi 3 (ARMCortex-A7 processor) and OP-TEE to compile the application and port it to Raspberry Pi. Some experience with embedded systems will be ideal for this project.
We will evaluate the utility privacy tradeoff of this application based on differential privacy aspects.
After completion of this project, you will learn how to use the trusted zones present in modern processors. With the ever growing concerns regarding privacy, the trust zone will be heavily used in the future. We will provide you with initial tutorials, the necessary hardware and software, in addition we offer a nice and friendly working atmosphere and a nice coffee machine ;).
Vaibhav Kulkarni (firstname.lastname at unil dot ch)