Attack of the Tails: Yes, You Really Can Backdoor Federated Learning

Abstract

Due to its decentralized nature, Federated Learning (FL) lends itself to adversarial attacks in the form of backdoors during training. The goal of a backdoor is to corrupt the performance of the trained model on specific sub-tasks (e.g., by classifying green cars as frogs). A range of FL backdoor attacks have been introduced in the literature, but also methods to defend against them, and it is currently an open question whether FL systems can be tailored to be robust against backdoors. In this work, we provide evidence to the contrary. We first establish that, in the general case, robustness to backdoors implies model robustness to adversarial examples, a major open problem in itself. Furthermore, detecting the presence of a backdoor in a FL model is unlikely assuming first order oracles or polynomial time. We couple our theoretical results with a new family of backdoor attacks, which we refer to as edge-case backdoors. An edge-case backdoor forces a model to misclassify on seemingly easy inputs that are however unlikely to be part of the training, or test data, i.e., they live on the tail of the input distribution. We explain how these edge-case backdoors can lead to unsavory failures and may have serious repercussions on fairness, and exhibit that with careful tuning at the side of the adversary, one can insert them across a range of machine learning tasks (e.g., image classification, OCR, text prediction, sentiment analysis).

References in corpus (31)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Communication-Efficient Learning of Deep Networks from Decentralized Data
• Federated Learning: Strategies for Improving Communication Efficiency
• Wide Residual Networks
• Understanding Black-box Predictions via Influence Functions
• Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples
• Federated Learning for Mobile Keyboard Prediction
• Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
• A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
• On the Convergence of FedAvg on Non-IID Data
• BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
• Towards Federated Learning at Scale: System Design
• Robust Aggregation for Federated Learning
• How To Backdoor Federated Learning
• Learning Differentially Private Recurrent Language Models
• Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates
• Differentially Private Federated Learning: A Client Level Perspective
• Applied Federated Learning: Improving Google Keyboard Query Suggestions
• Can You Really Backdoor Federated Learning?
• Local SGD Converges Fast and Communicates Little
• Practical Secure Aggregation for Federated Learning on User-Held Data
• From Parity to Preference-based Notions of Fairness in Classification
• An Overview of Machine Teaching
• Stronger Data Poisoning Attacks Break Data Sanitization Defenses
• The Non-IID Data Quagmire of Decentralized Machine Learning
• Zeno: Distributed Stochastic Gradient Descent with Suspicion-based Fault-tolerance
• Parallel SGD: When does averaging help?
• A Federated Learning Framework for Healthcare IoT devices
• Election Coding for Distributed Learning: Protecting SignSGD against Byzantine Attacks
• Teacher Improves Learning by Selecting a Training Subset

Cited by in corpus (17)

• FedML: A Research Library and Benchmark for Federated Machine Learning
• Group Knowledge Transfer: Federated Learning of Large CNNs at the Edge
• DeepSight: Mitigating Backdoor Attacks in Federated Learning Through Deep Model Inspection
• Machine Learning for Security in Vehicular Networks: A Comprehensive Survey
• Mime: Mimicking Centralized Stochastic Algorithms in Federated Learning
• Can Adversarial Weight Perturbations Inject Neural Backdoors?
• ScaleSFL: A Sharding Solution for Blockchain-Based Federated Learning
• Resisting Backdoor Attacks in Federated Learning via Bidirectional Elections and Individual Perspective
• Robust Distributed Optimization With Randomly Corrupted Gradients
• You Can Backdoor Personalized Federated Learning
• Active Poisoning: Efficient Backdoor Attacks on Transfer Learning-Based Brain-Computer Interfaces
• Byzantine-Robust Learning on Heterogeneous Datasets via Bucketing
• FLAIN: Mitigating Backdoor Attacks in Federated Learning via Flipping Weight Updates of Low-Activation Input Neurons
• Towards Bidirectional Protection in Federated Learning
• Defending Against Diverse Attacks in Federated Learning Through Consensus-Based Bi-Level Optimization
• Security and Privacy Issues and Solutions in Federated Learning for Digital Healthcare
• Semantic Host-free Trojan Attack