KO
|
EN
gitlite — search
Search
#python
#javascript
#typescript
#deep-learning
#nextjs
#react
#html
#php
#tailwindcss
#css
#flutter
#python3
VFL-Survey
★ 37
Open GitHub ↗
No description available.
Download README (.md)
Explore Similar Repositories
pFedGP
:
Code for Personalized Federated Learning with Gaussian Processes
af_backprop
:
No description available.
csapp-sol
:
深入理解计算机系统(第三版)课后习题+实验
DstServerHelper
:
自动处理饥荒联机版(Don't Starve Together)服务器的游戏更新、模组更新、崩溃、重置等,为服主分忧。
sc2-benchmark
:
[TMLR] "SC2 Benchmark: Supervised Compression for Split Computing"
// repository documentation
Was this content helpful?
★ 0
(0 ratings)
Select Rating:
★
★
★
★
★
Submit Feedback
Recent Feedback
×
Download README
Do you want to download the
README.md
file for
VFL-Survey
?
Download (.md)
# 纵向联邦学习算法及其隐私安全性综述 这个 Github 存储库总结了垂直联合学习资源的精选列表。 有关更多详细信息和分类标准,请参阅我们的综述论文。 什么是纵向联邦学习? 联邦学习(Federated Learning, FL)是一个机器学习框架,能有效帮助多个机构在满足数据隐私保护的要求下,利用各自的数据使用机器学习进行联合建模。FL旨在实现利用传输中间结果来实现学习目标,从而保证机构的原始数据都存储在本地,不进行任何的交换和传输。纵向联邦学习根据参与训练的机构的数据具有不同的特征分布,通常分为横向联邦学习(Horizontal FL, HFL)和纵向联邦学习(Vertical FL, VFL)。其中VFL适用于机构具有相同样本空间但不同特征空间的场景。近年来VFL被广泛应用于医疗健康、风险评估和广告推荐等领域。 1. 纵向联邦学习算法 - 1.1 基本模型 - 1.2 通信效率 - 1.3 数据对齐 - 1.4 标签分布 2. 纵向联邦学习的安全和隐私风险 - 2.1 标签推断攻击 - 2.2 属性推断攻击 - 2.3 成员推断攻击 - 2.4 数据重构攻击 - 2.5 后门攻击 3. 纵向联邦学习的防御 - 3.1 基于加密的防御 - 3.2 基于扰动的防御 - 3.3 基于对抗的防御 # 纵向联邦学习算法 ## 基本模型 Parallel Distributed Logistic Regression for Vertical Federated Learning without Third-Party Coordinator [[pdf]](https://arxiv.org/abs/1911.09824) Privacy Preserving Vertical Federated Learning for Tree-based Models [[pdf]](http://www.vldb.org/pvldb/vol13/p2090-wu.pdf) SecureBoost: A Lossless Federated Learning Framework [[pdf]](https://arxiv.org/abs/1901.08755) Split learning for health: Distributed deep learning without sharing raw patient data [[pdf]](https://arxiv.org/pdf/1812.00564.pdf) Multi-Participant Multi-Class Vertical Federated Learning [[pdf]](https://arxiv.org/abs/2001.11154) Learning Privately over Distributed Features: An ADMM Sharing Approach [[pdf]](https://arxiv.org/abs/1907.07735) A Vertical Federated Learning Framework for Graph Convolutional Network [[pdf]](https://arxiv.org/abs/2106.11593) Vertically Federated Graph Neural Network for privacy-preserving node classification [[pdf]](https://arxiv.org/abs/2005.11903) Federated Doubly Stochastic Kernel Learning for Vertically Partitioned Data [[pdf]](https://arxiv.org/abs/2008.06197) ## 通信效率 A Communication-Efficient Collaborative Learning Framework for Distributed Features [[pdf]](https://arxiv.org/abs/1912.11187) Private federated learning on vertically partitioned data via entity resolution and additively homomorphic encryption [[pdf]](https://arxiv.org/abs/1711.10677) A Quasi-Newton Method Based Vertical Federated Learning Framework for Logistic Regression [[pdf]](https://arxiv.org/abs/1912.00513) AsySQN: Faster Vertical Federated Learning Algorithms with Better Computation Resource Utilization [[pdf]](https://dl.acm.org/doi/abs/10.1145/3447548.3467169) Communication and Computation Reduction for Split Learning using Asynchronous Training [[pdf]](https://arxiv.org/abs/2107.09786) Privacy-Preserving Asynchronous Federated Learning Algorithms for Multi-Party Vertically Collaborative Learning [[pdf]](https://arxiv.org/abs/2008.06233) Model Optimization Method Based on Vertical Federated Learning [[pdf]](https://ieeexplore.ieee.org/document/9401521/) VAFL: a Method of Vertical Asynchronous Federated Learning [[pdf]](https://arxiv.org/abs/2007.06081) Cross-Silo Federated Learning for Multi-Tier Networks with Vertical and Horizontal Data Partitioning [[pdf]](https://arxiv.org/abs/2108.08930) Efficient Asynchronous Vertical Federated Learning via Gradient Prediction and Double-End Sparse Compression [[pdf]](https://ieeexplore.ieee.org/document/9305383) A Vertical Federated Learning Method for Interpretable Scorecard and Its Application in Credit Scoring [[pdf]](https://arxiv.org/abs/2009.06218) Accelerating Intra-Party Communication in Vertical FederatedLearning with RDMA [[pdf]](https://dl.acm.org/doi/10.1145/3426745.3431333) ## 对齐方式 Exploiting Record Similarity for Practical Vertical Federated Learning [[pdf]](https://arxiv.org/abs/2106.06312) FedMVT: Semi-supervised Vertical Federated Learning with MultiView Training [[pdf]](https://arxiv.org/abs/2008.10838) Asymmetrical Vertical Federated Learning [[pdf]](https://arxiv.org/abs/2004.07427) ## 标签分布 Secure Bilevel Asynchronous Vertical Federated Learning with Backward Updating [[pdf]](https://arxiv.org/abs/2103.00958) Multi-VFL: A Vertical Federated Learning System for Multiple Data and Label Owners [[pdf]](https://arxiv.org/abs/2106.05468) A Vertical Federated Learning Framework for Horizontally Partitioned Labels [[pdf]](https://arxiv.org/abs/2106.10056) # 纵向联邦学习的安全和隐私风险 ## 标签推断攻击 Label Inference Attacks Against Vertical Federated Learning [[pdf]](https://nesa.zju.edu.cn/download/fc_pdf_label_infer.pdf) Label Leakage and Protection in Two-party Split Learning [[pdf]](https://arxiv.org/abs/2102.08504) [[code]](https://github.com/bytedance/fedlearner/tree/master/example/privacy/label_protection) ## 属性推断攻击 Feature Inference Attack on Model Predictions in Vertical Federated Learning [[pdf]](https://arxiv.org/abs/2010.10152) [[code]](https://github.com/xj231/featureinference-vfl) Privacy-Preserving Federated Learning on Partitioned Attributes [[pdf]](https://arxiv.org/abs/2104.14383) Defending against Reconstruction Attack in Vertical Federated Learning [[pdf]](https://arxiv.org/abs/2107.09898) ## 成员推断攻击 Vertical Federated Learning without Revealing Intersection Membership [[pdf]](https://arxiv.org/abs/2106.05508) Asymmetric Private Set Intersection with Applications to Contact Tracing and Private Vertical Federated Machine Learning [[pdf]](https://arxiv.org/abs/2011.09350) [[code]](https://github.com/OpenMined/PSI) ## 数据重构攻击 Unleashing the Tiger: Inference Attacks on Split Learning [[pdf]](https://arxiv.org/abs/2012.02670) [[code]](https://github.com/pasquini-dario/SplitNN_FSHA) Understanding Deep Image Representations by Inverting Them [[pdf]](https://arxiv.org/abs/1412.0035) [[code]](https://github.com/novice03/timm-vis) Inverting Gradients - How easy is it to break privacy in federated learning? [[pdf]](https://arxiv.org/abs/2003.14053) [[code]](https://github.com/JonasGeiping/invertinggradients) See through Gradients: Image Batch Recovery via GradInversion [[pdf]](https://arxiv.org/abs/2104.07586) ## 后门攻击 Backdoor attacks and defenses in feature-partitioned collaborative learning [[pdf]](https://arxiv.org/abs/2007.03608) # 纵向联邦学习的防御 ## 基于加密的防御 A homomorphic-encryption-based vertical federated learning scheme for rick management [[pdf]](http://www.doiserbia.nb.rs/Article.aspx?id=1820-02142000022O#.YU_uuqh7rIU) Private federated learning on vertically partitioned data via entity resolution and additively homomorphic encryption [[pdf]](https://arxiv.org/abs/1711.10677) A Vertical Federated Learning Framework for Graph Convolutional Network [[pdf]](https://arxiv.org/abs/2106.11593) Vertically Federated Graph Neural Network for privacy-preserving node classification [[pdf]](https://arxiv.org/abs/2005.11903) ## 基于扰动的防御 Label Leakage and Protection in Two-party Split Learning [[pdf]](https://arxiv.org/abs/2102.08504) [[code]](https://github.com/bytedance/fedlearner/tree/master/example/privacy/label_protection) Hybrid Differentially Private Federated Learning on Vertically Partitioned Data [[pdf]](https://arxiv.org/abs/2009.02763) ## 基于对抗训练的防御 Defending against Reconstruction Attack in Vertical Federated Learning [[pdf]](https://arxiv.org/abs/2107.09898) Privacy-Preserving Federated Learning on Partitioned Attributes [[pdf]](https://arxiv.org/abs/2104.14383) # 纵向联邦学习平台 ## FATE FATE由微众银行开发,其提供了在横向、纵向和联邦迁移学习模式下实施联邦学习的框架。它可以通过docker镜像或手动步骤实现。该框架提供了具有Kubernetes集成的生产准备API。应⽤场景和技术实现,包括⻋险定价、信贷⻛控、销量预测、广告投放、⾃动驾驶、辅助诊断、视觉安防等。 [[Web]](https://github.com/FederatedAI/FATE) ## PaddleFL PaddleFL是基于百度PaddlePaddle开发的开源联邦学习框架。其支持横向联邦学习和纵向联邦学习。研究人员可以很轻松地用PaddleFL复制和比较不同的联邦学习算法,开发人员也比较容易在大规模分布式集群中部署PaddleFL联邦学习系统。PaddleFL提供很多种联邦学习策略(横向联邦学习、纵向联邦学习)及其在计算机视觉、自然语言处理、推荐算法等领域的应用。此外,PaddleFL还将提供传统机器学习训练策略的应用,例如多任务学习、联邦学习环境下的迁移学习。依靠着PaddlePaddle的大规模分布式训练和Kubernetes对训练任务的弹性调度能力,PaddleFL可以基于全栈开源软件轻松地部署。 [[Web]](https://github.com/PaddlePaddle/PaddleFL) ## Pysyft Pysyft是由OpenMined社区开发的基于安全和隐私的深度学习开源库。它在主流深度学习框架(例如PyTorch和TensorFlow)中使用联邦学习,差分隐私和加密计算(例如多方计算(MPC)和同态加密(HE)),将隐私数据与模型训练分离。在PyVertical中使用私有集交集 (PSI) 的私有实体解析,使用分裂网络( SplitNN)在垂直分区数据上训练模型。神经网络 (NN) 的训练在两个或多个参与方中进行拆分。每个参与方都拥有原始模型层的一个子集,即本地模型。每个参与方训练他们的本地模型,获得中间嵌入层,并将中间嵌入层发送到协作方。这允许在训练过程中在计算能力方面提高分裂神经网络的效率,同时保证性能。 [[Web]](https://github.com/OpenMined/PySyft) ## TF Encrypted TF-Encrypted(TFE)是用于安全多方计算 (MPC) 机器学习的开源框架。它建立在TensorFlow之上,利用了Keras API的易用性,同时通过安全的多方计算和同态加密对加密数据进行训练和预测。TF Encrypted的目的是让用户在并不精通密码学、分布式系统或高性能计算专业知识的情况下,也可以使用隐私保护加持的机器学习随时可用。用TF Encrypted的安全协议来安全汇总更新内容。当前实现使用了TF Encrypted和额外的密匙共享方法,需要三方(至更多方)实现安全地聚合计算。在安全计算中,其通过在图形模式下运行,并运用分发引擎按恰当顺序执行操作,实现同步功能。 [[Web]](https://github.com/tf-encrypted/tf-encrypted) ## CrypTen CrypTen是一个基于PyTorch的隐私保护机器学习框架。它目前将Secure Multiparty Computation作为其安全计算后端来实现。该框架通过CrypTensor外观与感觉完全像PyTorch的对象来呈现协议Tensor。这使用户可以使用类似于PyTorch中的自动区分和神经网络模块。CrypTen基于库。就像PyTorch一样,它实现了张量库。这使从业人员更容易调试。CrypTen 目前在Linux 和Mac上使用 Python 3.7运行。不支持Windows。目前尚不支持在GPU 上进行计算。 [[Web]](https://github.com/facebookresearch/CrypTen) ## iBond 智邦iBond是由同盾科技创建的一个分布式机器学习平台。该平台融合了分布式机器学习,安全加密计算,元学习等技术打造知识联邦,使多方联邦在完全满足用户隐私、数据安全和政府合规的要求下,进行数据分析和建模,协同创造和共享知识。该平台支持在原始数据的密文空间上联邦、模型训练中联邦、特征学习结果上进行联邦,还支持多任务、多方异构知识联邦。 [[Web]](https://www.tongdun.cn/) ## PowerFL PowerFL是由腾讯研发的金融级安全联合计算平台。其基于多数据源联合计算技术,提供稳定、高性能的联邦机器学习。构建在 Angel 机器学习平台上,利用 Angel-PS 支持万亿级模型训练的能力,将很多在 Worker 上的计算提升到参数服务器端;Angel PowerFL 为联邦学习算法提供了计算、加密、存储、状态同步等基本操作接口,通过流程调度模块协调参与方任务执行状态,而通信模块完成了任务训练过程中所有数据的传输。Angel PowerFL 联邦学习已经在腾讯金融云、腾讯广告联合建模等业务中开始落地。 [[Web]](https://data.qq.com/powerfl/)