Accelerated Vertical Federated Adversarial Learning through Decoupling Layer-Wise Dependencies
[NeurIPS 2025] [OpenReview] [Homepage] [PDF] [Code]
Vertical Federated Learning (VFL) enables participants to collaboratively train models on aligned samples while keeping their heterogeneous features private and distributed.
Despite their utility, VFL models remain vulnerable to adversarial attacks during inference.
Adversarial Training (AT), which generates adversarial examples at each training iteration, stands as the most effective defense for improving model robustness.
However, applying AT in VFL settings (VFAL) faces significant computational efficiency challenges, as the distributed training framework necessitates iterative propagations across participants.
To this end, we propose DecVFAL framework, which substantially accelerates VFAL training through a dual-level Decoupling mechanism applied during adversarial sample generation.
Specifically, we first decouple the bottom modules of clients (directly responsible for adversarial updates) from the remaining networks, enabling efficient lazy sequential propagations that reduce communication frequency through delayed gradients.
We further introduce decoupled parallel backpropagation to accelerate delayed gradient computation by eliminating idle waiting through parallel processing across modules.
Additionally, we are the first to establish convergence analysis for VFAL, rigorously characterizing how our decoupling mechanism interacts with existing VFL dynamics, and prove that DecVFAL achieves an
sudo apt update sudo apt install --upgrade screen tree vim git htop gcc g++ colordiff python-is-python3 byobu -y # sudo apt upgrade # sudo apt autoremove mkdir -p ~/miniconda3 wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3 rm ~/miniconda3/miniconda.sh source ~/miniconda3/bin/activate conda init --all conda update -n base -c defaults conda # a a
git clone https://github.com/workelaina/DecVFAL.git
# scp -r exp user@ip:~/vim ~/.bashrcalias ls="ls --color=auto" alias la="ls --color=auto -al" alias l="ls --color=auto -ahlF" alias diff='colordiff' alias grep='grep --color=auto' alias egrep='egrep --colour=auto' alias fgrep='fgrep --colour=auto' alias dua="du -sh *" alias vi='vim' alias py="python3" # export HF_HOME=/data/home # export HF_DATASETS_CACHE=/data/dataset # export TRANSFORMERS_CACHE=/data/tf # sudo chmod -R 777 /data conda deactivate conda activate decvfal # cd ~/exp
# torch 2.2.1 # cuda 12.1.1 # python 3.11 # conda clean -i # conda update -n base -c defaults conda conda create -y -n decvfal python=3.11 conda activate decvfal which python # https://pytorch.org/ # conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pytorch-cuda=12.1 -c pytorch -c nvidia -y # conda install pillow matplotlib numpy tqdm pandas scikit-learn scipy -c pytorch -c nvidia -y conda install pytorch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 pillow matplotlib numpy==1.26.4 tqdm pandas scikit-learn scipy pytorch-cuda=12.1 -c pytorch -c nvidia -y
nvidia-smi # which python python -c 'import torch;print(torch.cuda.is_available())' # lscpu # lsmem # lspci uname -a
screen -R sc1 # cd exp python train.py # python eval.py # Ctrl A D
@inproceedings{Accelerated2025TianxingMan, author = {Tianxing, Man and Yu, Bai and Ganyu, Wang and Jinjie, Fang and Haoran, Fang and Bin, Gu and Yi, Chang}, title = {Accelerated Vertical Federated Adversarial Learning through Decoupling Layer-Wise Dependencies}, year = {2025}, publisher = {Curran Associates Inc.}, booktitle = {Proceedings of the 39th International Conference on Neural Information Processing Systems}, series = {NIPS '25} }