Byzantine Robust Method

Byzantine robust methods aim to secure distributed machine learning, particularly federated learning (FL), against malicious actors ("Byzantine" nodes) that may inject faulty data or disrupt the training process. Current research focuses on developing algorithms that are robust to these attacks while maintaining efficiency, often employing techniques like gradient normalization, clipping, or multi-filter aggregation to identify and mitigate malicious inputs, even under non-identical data distributions and partial client participation. These advancements are crucial for ensuring the reliability and security of FL systems, which are increasingly important for privacy-preserving AI applications and large-scale distributed training.

Papers

September 26, 2024

Byzantine-Robust Aggregation for Securing Decentralized Federated Learning
Diego Cajaraville-Aboy, Ana Fernández-Vilas, Rebeca P. Díaz-Redondo, Manuel Fernández-Veiga
Decentralized Federated Learning Byzantine Robust Aggregation Byzantine Robust Method

August 18, 2024

Byzantine-resilient Federated Learning Employing Normalized Gradients on Non-IID Datasets
Shiyuan Zuo, Xingrun Yan, Rongfei Fan, Li Shen, Puning Zhao, Jie Xu, Han Hu
Data Set Non Convex Loss Function Byzantine Robust Federated Learning Federated Learning Approach Optimality Gap Federated Stochastic Byzantine Robust Method

May 14, 2024

Byzantine-Resilient Secure Aggregation for Federated Learning Without Privacy Compromises
Yue Xia, Christoph Hofmeister, Maximilian Egger, Rawad Bitar
Calysto Scheme Privacy Preserving Computation Secret Sharing Byzantine Robust Aggregation Byzantine Robust Method

December 2, 2023

PROFL: A Privacy-Preserving Federated Learning Method with Stringent Defense Against Poisoning Attacks
Yisheng Zhong, Li-Ping Wang
Federated Learning Poisoning Attack Homomorphic Encryption Byzantine Robust Federated Learning Byzantine Robust Method

November 23, 2023

Byzantine Robustness and Partial Participation Can Be Achieved at Once: Just Clip Gradient Differences
Grigory Malinovsky, Peter Richtárik, Samuel Horváth, Eduard Gorbunov
Machine Learning Model Variance Reduction Byzantine Robust Client Sampling Gradient Clipping Byzantine Fault Tolerance Partial Participation Byzantine Robust Method

November 8, 2023

Byzantine-Tolerant Methods for Distributed Variational Inequalities
Nazarii Tupitsa, Abdulla Jasem Almansoori, Yanlin Wu, Martin Takáč, Karthik Nandakumar, Samuel Horváth, Eduard Gorbunov
Byzantine Robust Variational Inequality Byzantine Attack Byzantine Robust Method

October 15, 2023

Communication Compression for Byzantine Robust Learning: New Efficient Algorithms and Improved Rates
Ahmad Rammal, Kaja Gruntkowska, Nikita Fedin, Eduard Gorbunov, Peter Richtárik
Novel Algorithm Byzantine Robust Fast Rate Communication Compression Smooth Convex Optimization Byzantine Robust Method

March 9, 2023

FedREP: A Byzantine-Robust, Communication-Efficient and Privacy-Preserving Framework for Federated Learning
Yi-Rui Yang, Kun Wang, Wu-Jun Li
New Framework Byzantine Robust Byzantine Robust Method

June 1, 2022

Variance Reduction is an Antidote to Byzantines: Better Rates, Weaker Assumptions and Communication Compression as a Cherry on the Top
Eduard Gorbunov, Samuel Horváth, Peter Richtárik, Gauthier Gidel
Loss Function Variance Reduction Byzantine Robust Fast Rate Communication Compression Byzantine Seal Byzantine Robust Method

Byzantine Robust Method

Papers

Byzantine-Robust Aggregation for Securing Decentralized Federated Learning

Byzantine-resilient Federated Learning Employing Normalized Gradients on Non-IID Datasets

Byzantine-Resilient Secure Aggregation for Federated Learning Without Privacy Compromises

PROFL: A Privacy-Preserving Federated Learning Method with Stringent Defense Against Poisoning Attacks

Byzantine Robustness and Partial Participation Can Be Achieved at Once: Just Clip Gradient Differences

Byzantine-Tolerant Methods for Distributed Variational Inequalities

Communication Compression for Byzantine Robust Learning: New Efficient Algorithms and Improved Rates

FedREP: A Byzantine-Robust, Communication-Efficient and Privacy-Preserving Framework for Federated Learning

Variance Reduction is an Antidote to Byzantines: Better Rates, Weaker Assumptions and Communication Compression as a Cherry on the Top