UPSI Revisited

Updatable Private Set Intersection: Extended Functionalities, Deletion, and Worst-Case Complexity

Saikrishna Badrinarayanan, Peihan Miao, Xinyi Shi, Max Tromanhauser, and Ruida Zeng

Overview | Building | Experiments | Contact | License | Citation

---

Overview

UPSI Revisited implements the Updatable Private Set Intersection (UPSI) protocol as detailed in our Asiacrypt 2024 paper. Private Set Intersection (PSI) enables two mutually distrusting parties, each holding a private set of elements, to compute the intersection of their sets without disclosing any additional information. Building upon the foundational work presented in PoPETS 2022, our UPSI Revisited project addresses several key limitations of existing UPSI protocols:

1. Extended Functionalities: Unlike previous protocols that support only plain PSI, our implementation includes advanced functionalities such as PSI-Cardinality and PSI-Sum. Additionally, in the addition-only setting, we present Circuit-PSI functionality that outputs secret shares of the intersection.
2. Support for Deletion Operations: Previous UPSI protocols were limited to the addition of elements to their existing sets and "weak deletions" (where parties can additionally delete their old elements every t days). Our work introduces the capability to arbitrarily delete elements, achieving semi-honest security in both the addition-only and addition-deletion settings.
3. Optimized Worst-Case Complexity: Existing addition-only protocols either require both parties to learn the output or only achieve low amortized complexity and incur linear worst-case complexity. Our protocols ensure that both computation and communication complexities scale solely with the size of set updates rather than the entire sets (except for a polylogarithmic factor).

Practical Performance: We have implemented our UPSI protocols and benchmarked them against state-of-the-art PSI and extended functionality protocols. Our results demonstrate favorable performance, particularly when dealing with sufficiently large total set sizes, sufficiently small new updates, or operating within low-bandwidth network environments.

Warning

This repository is a research prototype written to demonstrate our UPSI protocol's performance and to showcase its capabilities. It is NOT intended to be considered as "production ready" and should only be used for experimental or research & development purposes.

Building the Project

Container Setup with Docker

The repository has been containerized using Docker. To pull the appropriate container:

docker pull ghcr.io/ruidazeng/upsi-revisited:latest

Standard Environments

If you're using a standard environment (e.g. amd64 Intel-based systems), run:

docker run -it ghcr.io/ruidazeng/upsi-revisited:latest

Warning

If you run into an error RTNETLINK answers: Operation not permitted, add --cap-add=NET_ADMIN to the docker run command.

Running the Container on Apple Silicon Macs or ARM Architectures

For users operating on Apple Silicon Macs (arm64) or other ARM-based architectures, it might be necessary to specify the platform explicitly to direct Docker to emulate the standard environments and architectures, ensure compatibility and allowing the container to run:

docker run --platform linux/amd64 -it ghcr.io/ruidazeng/upsi-revisited:latest

Once inside the container, you can run any of the commands outlined in the following sections.

Building Locally

If you don't want to use Docker, you can build the project yourself.

Important

The project should be built on Debian 10 (buster) to ensure dependencies will build without issue.

The build.sh script will handle installation of build dependencies (namely python, bazel, and emp-toolkit). The UPSI library can then be built using the bazel endpoints:

# for the addition only cardinality, sum, and circuit updatable psi protocols
bazel build //upsi/addition:all

# for the addition & deletion cardinality and sum updatable psi protocols
bazel build //upsi/deletion:all

# for the addition only plain psi protocol
bazel build //upsi/original:all

# for the addition & deletion plain psi protocol
bazel build //upsi/deletion-psi:all

Running the Experiments

Before running experiments, use the setup binary to generate encryption keys and mock input sets. By default, keys and input sets will be placed in out/ and data/, respectively.

Protocol Descriptions

The protocols correspond to specific functionalities as outlined in the paper:

• addition: Addition-Only UPSI-Cardinalty/Sum/Circuit-PSI
• deletion: UPSI-Cardinalty/Sum with Addition and Deletion
• original: Addition-Only Plain UPSI
• deletion-psi: Plain UPSI with Addition and Deletion

For more details, refer to the corresponding sections in the paper:

• addition & original: Section 3.2
• deletion & deletion-psi: Section 4.2

To explore the parameters for each protocol, use the --help flag for the setup binary as follows:

./bazel-bin/upsi/<protocol>/setup --help

The --func flag specifies the functionality to be used in the experiment. The available values are:

• PSI: Plain PSI, allowing two parties to compute the intersection of their sets without revealing any additional information.

• CA: PSI-Cardinality, which computes the size of the intersection between the two sets.

• SUM: PSI-Sum, calculating the sum of values associated with the elements in the intersection.

• SS: Circuit-PSI, outputting the secret shares (SS) of the intersection.

To replicate the fourth row in Table 2 ($N = 2^{18}$, $N_d = 2^6$ running the updatable PSI addition only for cardinality protocol $\Pi_{\mathsf{UPSI-Add}_\mathsf{ca}}$), we want to run the protocol on the day where the input sets reach total cardinality of $2^{18}$ (i.e., the $\frac{2^{18}}{2^{6}} = 4096^\text{th}$ day). Rather than simulate all 4096 days, specify the --start_size parameter in the setup binary to generate encrypted datasets that are used as the "carry over" from the $4095^\text{th}$ to $4096^\text{th}$ day. Therefore, to set up this experiment use:

./bazel-bin/upsi/addition/setup --func=CA --days=1 --daily_size=64 --start_size=262080

Note that $262080 = 2^{18} - 2^6 = N - N_d$; i.e., the size of the input sets at the start of the $4096^\text{th}$ day.

Configuring Network Bandwidth and Latency for Experiments

To simplify the setup of network conditions for experiments, the network_setup.sh script is provided in the base directory. This script automates the configuration of network bandwidth and latency, simulating both LAN and WAN environments as described in Section 6.1 of the paper.

Overview of Network Simulations

In Section 6.1, the paper explains the network conditions used for experiments:

• LAN Connection:
  • RTT (Round Trip Time): 0.2 ms
  • Bandwidth: 1 Gbps
• WAN Connection:
  • RTT (Round Trip Time): 80 ms
  • Bandwidth Options: 200 Mbps, 50 Mbps, and 5 Mbps

These settings follow the same settings as previous works and are critical for reproducing the results in Tables 2, 3, 4, and 5.

Usage of network_setup.sh

1. Enable a Specific Network Setting: Use the following command to set up a specific network condition. Replace <latency> with the RTT latency in milliseconds and <bandwidth> with the desired (optional) max bandwidth in Mbps.

   ./network_setup.sh on <latency> <bandwidth>

   Example configurations:

   • LAN (1 Gbps, 0.2 ms RTT):

     ./network_setup.sh on 0.2 1000

   • WAN (200 Mbps, 80 ms RTT):

     ./network_setup.sh on 80 200

   • WAN (50 Mbps, 80 ms RTT):

     ./network_setup.sh on 80 50

   • WAN (5 Mbps, 80 ms RTT):

     ./network_setup.sh on 80 5

2. Disable Network Emulation: After completing the experiments under the specified network condition, disable the network emulation by running:

   ./network_setup.sh off

Example Workflow for Experiments

Once the encryption keys and data sets are generated, and the network is setup as desired, each party in the protocol can be run:

./bazel-bin/upsi/addition/run --party=1 --days=1 --func=CA

and

./bazel-bin/upsi/addition/run --party=0 --days=1 --func=CA

Note

You must run run --party=1 before run --party=0 as the first party will wait for connections on the specified port.

Or to run both parties, you can do:

./bazel-bin/upsi/addition/run --party=1 --days=1 --func=CA & ./bazel-bin/upsi/addition/run --party=0 --days=1 --func=CA

Important

If you've used --start_size in the setup command, you must include the --trees flag in run to read and use the prepared state.

Below is a step-by-step example workflow to configure and run experiments under LAN and WAN conditions:

1. Run the Experiment on LAN (1 Gbps):

   ./network_setup.sh on 0.2 1000
   ./bazel-bin/upsi/addition/run --party=1 --days=1 --func=CA
   ./bazel-bin/upsi/addition/run --party=0 --days=1 --func=CA
   ./network_setup.sh off

2. Run the Experiment on WAN (200 Mbps):

   ./network_setup.sh on 80 200
   ./bazel-bin/upsi/addition/run --party=1 --days=1 --func=CA
   ./bazel-bin/upsi/addition/run --party=0 --days=1 --func=CA
   ./network_setup.sh off

3. Run the Experiment on WAN (50 Mbps):

   ./network_setup.sh on 80 50
   ./bazel-bin/upsi/addition/run --party=1 --days=1 --func=CA
   ./bazel-bin/upsi/addition/run --party=0 --days=1 --func=CA
   ./network_setup.sh off

4. Run the Experiment on WAN (5 Mbps):

   ./network_setup.sh on 80 5
   ./bazel-bin/upsi/addition/run --party=1 --days=1 --func=CA
   ./bazel-bin/upsi/addition/run --party=0 --days=1 --func=CA
   ./network_setup.sh off

More information for both the setup and run binaries can be found using the --help flag.

Author Contact Information

Feel free to reach out to the authors for further inquiries or collaborations:

Name	Affiliation	Contact
Saikrishna Badrinarayanan	LinkedIn	bsaikrishna7393 [at] gmail [dot] com
Peihan Miao	Brown University	peihan_miao [at] brown [dot] edu
Xinyi Shi	Brown University	xinyi_shi [at] brown [dot] edu
Max Tromanhauser	Brown University	max_tromanhauser [at] brown [dot] edu
Ruida Zeng	Brown University	ruida_zeng [at] brown [dot] edu

License

This project is licensed under the Apache License 2.0.

To enable easy citation of your work, you can add a citation section like this:

Citation

If you find our work useful in your research or reference it in your publications, please consider citing it as follows:

@inproceedings{bmxtz2024upsi,
  title={{UPSI Revisited: Updatable Private Set Intersection: Extended Functionalities, Deletion, and Worst-Case Complexity}},
  author={Saikrishna Badrinarayanan and Peihan Miao and Xinyi Shi and Max Tromanhauser and Ruida Zeng},
  booktitle={Advances in Cryptology -- ASIACRYPT 2024},
  year={2024},
  publisher={Springer},
  url={https://eprint.iacr.org/2024/1446},
}