Before :

The Problem

• Queries can be only be performed on plain text
• Analysis, indexing, operations 
• Models need to read the raw data to process it
• A lot of computation is need it to decry data on the fly, search is a complex problem 

The Solution

Fully Homographic Encryption Using Ideal Lattices

Use Cases

• Extracting value from private data
• data set intersection
• gnocmics analitycs 
• oblivious queries
• secure computing outsourcing
• health care / finances records mining 

Capabilities

• Ability to perform aromatic operations on encrypted data
• No need to decrypt
• Query a database that is encrypted

Progression

Deep Dive into Homomorphic Encryption - THR3001 - Microsoft Research

High level workflow

An actual architecture

Real world validations 

Real world validations 

Real world validations 

What about images?

Real world validations 

Image Privacy Framework

Image Privacy Framework

A Practical Demo with PHE

Scheme Rationale

Demo & hands-on

• Using Partially HE (PHE) Paillier Crytosystem (Asymetric) in Python
• Using HElib C++ IMB Fully HE with BGV Scheme for Queries
• Using Microsoft SEAL with BFV in C++/Python

References

• Private queries on encrypted genomic data, DOI: 10.1186/s12920-017-0276-z
• IBM 
• Microsoft Research