SGX-PySpark Demo

This repository contains SGX-PySpark demo.

Video

SGX-PySpark demo

Details

First, please shoot us ([email protected]) an email providing your Docker ID to get access to the Docker image: sconecuratedimages/www2019:sgx-pyspark

Try it out by executing:

git clone https://github.com/doflink/sgx-pyspark-demo && cd sgx-pyspark-demo 

docker run -it --rm -v `pwd`:/fspf  --privileged -p 8080:8080 -p 6868:6868 -p 28778:28778 sconecuratedimages/www2019:sgx-pyspark sh

Go to the demo directory:

 cd /fspf/

Create a file system protection file (meta file) to store all the metadata required for checking the consistency of files. Then add encrypted regions and encrypt the input PySpark codes and data. Store the encryption key and the tag of the fspf to the file keytag:

 ./fspf.sh

Export environment variables

export SCONE_FSPF_KEY=$(cat input/keytag | awk '{print $11}')
export SCONE_FSPF_TAG=$(cat input/keytag | awk '{print $9}')
export SCONE_FSPF=/fspf/encrypted-files/volume.fspf

Now, run the wordcount application with SGX-PySpark.

/spark/bin/spark-submit encrypted-files/enc-wordcount.py  input/encrypted-sensitive-input.txt  spark://$(hostname -f):7077 &> output.txt &

Try to dump memory of the application to steal secrets

./memory-dump.sh

Contacts

Send email to [email protected] or [email protected]