PETSQL

PETSQL is a framework for secure collaborative data analysis based on Secure Multi-Party Computation (MPC), and its core capabilities are supported by PETAce. PETSQL integrates MPC with the Spark big data engine and allow users to initiate collaborative computing tasks using standard SQL statements. Designed specifically for big data privacy needs, the framework encapsulates data objects and computational operations. As a result, by enabling the framework to handle big data objects and tasks, we can seamlessly extend big data support without requiring any front-end modifications. PETSQL primarily comprises the following parts:

• Compiler and Transporter: A compiler parses SQL statements and schema information into a structured logical execution plan. A transporter takes into account the characteristics of MPC and rewrites the logical execution plan into real processes.

• Virtual Machine: A virtual machine defines a series of interfaces that allow plug-and-play implementations to cater to different scenarios. It mainly includes two separate engines for plaintext and ciphertext that correspond to operations on unencrypted and encrypted data, respectively.

• Data Layer: A data layer abstracts various types of data, such as in-memory and database storage. This is an underlying architecture of the virtual machine that hides away discrepancies between data storage and formats to maximize code reuse.

The diagram below shows the overall architecture of PETSQL and the relationships between its various modules.

Architecture of PETSQL

Getting Started

Requirements

System	Toolchain
Linux	Python (==3.9), pip(>=23.3.1) Java 8, Maven(>=3.6.0)

Building PETSQL

PETSQL depends on PETAce as the backend for secure multi-party computation. PETAce must be installed first. Please refer to PETAce.

We assume that all commands presented below are executed in the root directory of PETSQL. Then, build java using the following scripts:

cd java
mvn clean package
cd ..
cp ./java/target/petsql*.jar ./petsql/compiler/__binding/

Then, build python using the following scripts:

mkdir -p build/petsql
cp -r petsql/* build/petsql
cp setup.py build
cp requirements.txt build
cd build
python3 setup.py install

Quick Start

Here we give a simple example to run protocols in PETSQL.

To run Party A

python3 ./example/memory.py -p 0

To run Party B

python3 ./example/memory.py -p 1

For more details, please refer to the example.

Contribution

Please check Contributing for more details.

Code of Conduct

Please check Code of Conduct for more details.

License

This project is licensed under the Apache-2.0 License.