Efficient Secure Aggregation for Privacy-Preserving Federated Machine Learning

This repository contains the full implementation of e-SeaFL paper (accepted to ACSAC 2024).

Warning: This code is a proof-of-concept prototype and is not ready for production use.

Awards

This work has been awarded all three available badges—Artifact Available, Reviewed, and Reproducible—by the ACSAC 2024 reviewers.

Code Structure

• ServerCode/: Contains the Server.py file, which handles the server-side logic in the system.
• UserCode/: Contains the User.py file, which manages the operations of users participating in the system.
• AssistingNodeCode/: Contains the AssistingNodeCode.py file, which manages the operations of assisting nodes in the system.
• AesModeCTR-master/: Contains the C++ implementation for AES encryption in CTR mode.
• Compilation Scripts:
  • auto_setup.sh: Bash script to automatically install the required libraries and packages.
  • script.sh: Bash script to compile necessary binaries (aggregation.so, AesModeCTR) and run the entire system.

Prerequisites

Before running the system, ensure that the following software is installed:

• Python 3.8+
• GCC (for compiling C++ code)
• OpenSSL (for cryptographic operations)

How to Run

To automatically install the required libraries and packages, run the following script:

./auto_setup.sh

Running e-SeaFL Code

To run the e-SeaFL system, use the following command:

./script.sh -u <number_of_users> -a <number_of_assisting_nodes> -c <commitment> [-o <printAggOutput>] [-p <server_port>] [-b <bandwidth_mode>]

Parameters:

• -u <number_of_users>: Required. The number of users participating in the federated learning process. Must be greater than 2.
• -a <number_of_assisting_nodes>: Required. The number of assisting nodes in the system.
• -c <commitment>: Required. Set to 0 to run without commitment (i.e., without model integrity check) or `1` to run with commitment (i.e., with model integrity check).
• -o <printAggOutput>: Optional. Controls the output behavior:
  • 0: No print (default).
  • 1: Print result.
  • 2: Print summary.
• -b <bandwidth_mode>: Optional. Controls the output behavior for outbound bandwidth:
  • 0: No print (default).
  • 1: Print outbound bandwidth.
  • 2: Only print outbound bandwidth.
• -p <server_port>: Optional. The port number for the server (default is `9000`, valid range is `2000-60000`).

Examples:

Running a quick test with 25 users, 3 assisting nodes, without commitment, and printing the summary:

./script.sh -u 25 -a 3 -c 0 -o 2 -p 10050

Running a quick test with 20 users, 3 assisting nodes, with commitment, and printing the summary:

./script.sh -u 20 -a 3 -c 1 -o 2 -p 10070

Running with 400 users, 3 assisting nodes, and without commitment:

./script.sh -u 400 -a 3 -c 0 -p 10400

Running with 200 users, 3 assisting nodes, with commitment, and printing the summary:

./script.sh -u 200 -a 3 -c 1 -o 2 -p 10300

Running with 600 users, 3 assisting nodes, without commitment, and printing the outbound bandwidth:

./script.sh -u 600 -a 3 -c 0 -b 1 -p 15000

Artifact Documentation

For a detailed description of how to reproduce the results presented in the paper, please refer to our Artifact Documentation.

Citing

If you use this repository or build upon our work, we would appreciate it if you cite our paper using the following BibTeX entry:

@inproceedings{eSeaFL2024,
  author={Behnia, Rouzbeh and Riasi, Arman and Ebrahimi, Reza and Chow, Sherman S. M. and Padmanabhan, Balaji and Hoang, Thang},
  booktitle={2024 Annual Computer Security Applications Conference (ACSAC)}, 
  title={Efficient Secure Aggregation for Privacy-Preserving Federated Machine Learning}, 
  year={2024},
  volume={},
  number={},
  pages={778-793},
  doi={10.1109/ACSAC63791.2024.00069}}
}