trustworthycomputing / plasma

Code for "PLASMA: Private, Lightweight Aggregated Statistics against Malicious Adversaries"

PLASMA: Private, Lightweight Aggregated Statistics against Malicious Adversaries

WARNING: This is not production-ready code.

This is software for a research prototype. Please do NOT use this code in production. This repository builds upon heavy-hitters.

How to cite this work

PLASMA will appear in PoPETS 2024 (the preprint can be accessed here). You can cite this work as follows:

@Article{PoPETS:MouSarTso24,
  author =       "Dimitris Mouris and
                  Pratik Sarkar and
                  Nektarios Georgios Tsoutsos",
  title =        "{PLASMA: Private, Lightweight Aggregated Statistics against Malicious Adversaries}",
  year =         2024,
  volume =       2024,
  month =        July,
  journal =      "{Proceedings on Privacy Enhancing Technologies}",
  number =       3,
  pages =        "1--19",
}

Build & Run With Docker Compose

The following runs three servers and the leader each in a different container:

❯❯ docker compose up

Getting started

First, make sure that you have a working Rust installation:

❯❯ rustc --version
rustc 1.67.1
❯❯ cargo --version
cargo 1.67.1

Note that we don't require the exact Rust version, but this is the one we used in our experiments.

Build from sources

cargo build --release

Run the tests:

cargo test --release

Heavy Hitters

First off, run the servers in different shells.

Server 0:

cargo run --release --bin server -- --config src/bin/config_8.json --server_id 0

Server 1:

cargo run --release --bin server -- --config src/bin/config_8.json --server_id 1

Server 2:

cargo run --release --bin server -- --config src/bin/config_8.json --server_id 2

Now, the servers should be ready to process client requests. In a forth shell, run the following command to send 100 client requests to the servers:

Clients:

cargo run --release --bin leader -- --config src/bin/config_8.json -n 100

To run with the presence of malicious clients include the --malicious flag followed by the percentage of malicious clients to generate ([0.0, 0.9]). For instance, to run with 5% malicious clients use:

cargo run --release --bin leader -- --config src/bin/config_8.json -n 100 --malicious 0.05

The config file

The client and servers use a common configuration file, which contains the parameters for the system. An example of one such file is in src/bin/config_8.json. The contents of that file are here:

{
  "data_bytes": 1,
  "threshold": 0.01,
  "server_0": "0.0.0.0:8000",
  "server_1": "0.0.0.0:8001",
  "server_2": "0.0.0.0:8002",
  "addkey_batch_size": 1000,
  "hashes_batch_size": 100000,
  "unique_buckets": 1000,
  "zipf_exponent": 1.03
}

The parameters are:

• data_bytes: Number of bytes of each string (x8 for bits).
• threshold: The servers will output the collection of strings that more than a threshold of clients hold.
• server0, server1, and server2: The IP:port of tuple for the two servers. The servers can run on different IP addresses, but these IPs must be publicly addressable.
• addkey_batch_size: The number of each type of RPC request to bundle together. The underlying RPC library has an annoying limit on the size of each RPC request, so you cannot set these values too large.
• unique_buckets and zipf_exponent: Each simulated client samples its private string from a Zipf distribution over strings with parameter zipf_exponent and support unique_buckets.

Trustworthy Computing Group