Segam: Secure and Efficient Group-by-Aggregation Queries across Multiple Private Database

Features

Segam is a secure collaborative analytics system designed to execute secure group-by-aggregation queries. To safeguard privacy, we leverage the well-established secure multi-party computation technique to design secure operators for robust security guarantees. In brief, the features of Segam can summarize as follows:

Practical Secure Queries: Segam provides a practical approach to efficiently handle secure group-by-aggregation queries.
Secure Operators for Privacy: Segam decompose queries into multiple rounds of plaintext and secure operations and design various secure operators based on Shamir's secret sharing scheme to fortify the security foundation.
User-Friendly SQL Support: Segam supports SQL queries and offers a straightforward experience. Quick-start examples with Docker facilitate a smooth deployment process.

Requirements

Ubuntu 18.04
Docker
Java 11
Apache Maven 3.6.0+

Note: The system has been tested on Ubuntu, but theoretically, it is also applicable to other operating systems such as macOS and Windows.

Example

First, ensure that the above requirements are installed.
Granting Docker permissions to the account (optional). You can try running "docker run hello-world". If you don't see the message "Permission Denied", then there's no need to proceed with this step.

# Verify the existence of the Docker group
cat /etc/group | grep docker

# If the Docker group is not present in the output, please create the group.
sudo groupadd docker

# Add the current user to the Docker group.
sudo usermod -aG docker $USER

Clone the git repository

git clone git@github.com:system-segam/segam.git
cd segam

Deploying three data owners using Docker and initializing them.

docker pull postgres:13
# Package and deploy
chmod +x ./package.sh && ./package.sh
chmod +x ./deploy.sh && ./deploy.sh

You can verify whether the data container has been initialized successfully by executing the following command.

# If it displays the Java version and the number of rows in the data table correctly, it indicates successful initialization.

docker exec -u postgres container1 bash -c 'java --version && psql postgres -c "select count(*) from lineitem"'
docker exec -u postgres container2 bash -c 'java --version && psql postgres -c "select count(*) from lineitem"'
docker exec -u postgres container3 bash -c 'java --version && psql postgres -c "select count(*) from lineitem"'

Start up all data owner services.

cd release
./start-server.sh

Start up query side service.

# You can view the parameter instructions by using java -jar core.jar -h
java -jar core.jar -i test_sum.sql -p private

Run your own query

If you wish to query your own data, the following steps should be performed, along with modifications to certain configuration files as outlined below.

Build multiple Docker containers with configurations including Postgres 13+ and Java 11+, and map port 9991 of the container (which can be specified by modifying the serverPort item in container/config.json) to a custom port on the host machine.
Modify the release/config.json file to represent each container using the "local network IP of the host machine + the port number mapped to the host machine" obtained from the previous step.
Modify the release/container/config.json file to configure the access method for the corresponding container databases.

For scenarios where a real multi-machine environment is used instead of Docker containers, the overall steps are similar.

Considerations

This section will be continuously updated in the future.

Segam utilizes the gRPC framework, modifying **.proto files in segam-core/src/main/resources/ and executing the following command to generate code:

cd segam-core
call mvn protobuf:compile -f pom.xml
call mvn protobuf:compile-custom -f pom.xml
cd ..