CRFL: Certifiably Robust Federated Learning against Backdoor Attacks

Authors: Chulin Xie, Minghao Chen, Pin-Yu Chen, Bo Li

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Practically, we conduct comprehensive experiments across a range of federated datasets, and provide the first benchmark for certified robustness against backdoor attacks in federated learning.
Researcher Affiliation Collaboration 1University of Illinois at Urbana-Champaign 2Zhejiang University 3IBM Research. Correspondence to: Chulin Xie <chulinx2@illinois.edu>, Pin-Yu Chen <pinyu.chen@ibm.com>, Bo Li <lbo@illinois.edu>.
Pseudocode Yes Algorithm 1 Federated averaging with parameters clipping and perturbing; Algorithm 2 Certification of parameters smoothing
Open Source Code Yes Our code is publicaly available at https://github.com/AI-secure/CRFL.
Open Datasets Yes We train the FL system following our CRFL framework with three datasets: Lending Club Loan Data (LOAN) (Kan, 2019), MNIST (Le Cun & Cortes, 2010), and EMNIST (Cohen et al., 2017).
Dataset Splits No The paper mentions 'training data' and 'test sets' but does not explicitly provide percentages, sample counts, or specific predefined splits for training, validation, and testing datasets to reproduce the experiment.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., GPU models, CPU types, memory) used to run the experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers (e.g., Python version, library versions like PyTorch, TensorFlow, scikit-learn).
Experiment Setup Yes In all experiments, unless otherwise stated, we use σT = 0.01 to generate M = 1000 noisy models in parameter smoothing procedure, and use the error tolerance α = 0.001.