karannewatia/mycelium

Mycelium

This is a partial implementation of Mycelium, a framework for private graph analytics over massive distributed data. This is not a completely distributed end-to-end implementation (we cannot simulate the millions of users required!) It instead consists of microbenchmarks for user operations, aggregator operations, and elected committee operations.

Initial installation:

• git clone https://github.com/karannewatia/Mycelium
• cd Mycelium
• ./install.sh (requires python version v>3.8 and pip)

Summary map of repository contents:

crypto:

• Python code for user-side/aggregator-side FHE operations (crypto/code/python_version/)
• MPC files for decryption (crypto/code/scale_mamba_version/)
• Amazon EC2 experiment instructions/scripts (crypto/mpc_amazon_ec2_benchmarking/)

zkp

• code for ZK proofs

original_graphs

• the graphs we generated from the data we obtained

graph_scripts

• all scripts used for generating the graphs in the paper

re_sharing:

• Re-sharing (extended VSR protocol) proof-of-concept/rough implementation

onion-routing:

• all onion routing code

Instructions for Artifact Reviewers at SOSP 2021:

Our implementation is composed of a series of micro-benchmarks. Please cd into each of the above folders for detailed explanations and instructions on evaluating each component separately. We provide a docker for some of the cryptographic operations, but require access to CloudLab to evaluate our onion-routing costs. The remainder of operations can be evaluated without any specialized hardware or access.

We summarize what is required to replicate all empirical results below:

Replication of Graphs and Results in Paper

Figure 5

Follow README in graph_scripts/

Figure 6

Follow READMEs in zkp/ and onion-routing/ to generate all costs. Then generate graph in graph_scripts/

User Computation Costs (Section 6.4)

Follow READMEs in crypto/code/python_version/, onion-routing, and zkp/ to generate all costs. The user computation cost includes the costs of encrypting a ciphertext and multiplying 10 ciphertexts (run encryption_test.py and multiplication_test.py in crypto/code/python_version/), telescoping and sending a message (see onion-routing), and generating the ZK proofs for encryption and multiplication (see zkp/).

Committee Costs (Section 6.5)

Follow READMEs in crypto/code/scale_mamba_version/ for local simulation and/or single-machine EC2 simulation

Figure 8

Follow README in graph_scripts/

Figure 9

Follow READMEs in crypto/code/python_version/, onion-routing and zkp/ to generate all costs. Then generate graph in graph_scripts/

List of open-source libraries used

• ZoKrates
• OpenSSL
• SCALE-MAMBA