DBA: Distributed Backdoor Attacks against Federated Learning

By Chulin Xie, Keli Huang, Pin-Yu Chen, Bo Li. 2020

947 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Backdoor attacks and defenses in feature-partitioned collaborative learning (2020, background): “Recent works (Bagdasaryan et al., 2018; Xie et al., 2020) use differential privacy to defense the backdoor attacks in federated learning.”
Backdoor Learning: A Survey (2020, background): “[37] introduced a distributed backdoor attacks to federated learning.”
LIRA: Learnable, Imperceptible and Robust Backdoor Attacks (2021, background): “As this field of research has evolved, the strength and capabilities of these attacks have increased, leading to methodologies that work with stealthier triggers [51] or compromise extended scenarios [4, 53].”
Biscotti: A Blockchain System for Private and Secure Federated Learning (2021, methodology): “Prior work has demonstrated that adversaries can attack the shared model through poisoning attacks [11], [12], [15], [16], [17], [18], [19], in which an adversary contributes adversarial updates to shared model parameters.”
Gradient-Leaks: Understanding and Controlling Deanonymization in Federated Learning (2018, background): “One line of work studies malicious devices who exploit anonymity and secure aggregation protocols to mount poisoning and backdoor attacks [7,12,82] on the system.”
Dataset Security for Machine Learning: Data Poisoning, Backdoor Attacks, and Defenses (2020, unknown): “[117] proposes distributed backdoor attacks, which better”
A Comprehensive Survey of Privacy-preserving Federated Learning (2021, unknown): “WHO RA1 [61, 196, 228] [26, 137, 218] [157, 169, 198] [144, 164, 199] [63, 70, 193] [16, 174, 191] [78, 110, 228] [52, 81, 216] [32, 56, 222] [30, 127, 173] RA2 [114, 175, 187] [215, 221, 224] [29, 111, 113] [13, 199] [3, 10, 40] [109, 155, 168] [45, 172, 193] [20, 52, 57] [56, 165, 197] [173, 199, ”
FLAME: Taming Backdoors in Federated Learning (2022, methodology): “In this paper, we therefore focus on the more challenging problem of backdoor attacks [7, 45, 57, 59], i.e., targeted poisoning attacks in which the adversary seeks to stealthily manipulate the resulting global model in a way that attacker-controlled inputs result in incorrect predictions chosen by…”
Survey on Federated Learning Threats: concepts, taxonomy on attacks and defences, experimental study and challenges (2022, background): “It is composed of several coordinated adversarial clients (sybils), where each of them injects a different pattern or part of a common pattern to the model [42].”
Understanding and Controlling User Linkability in Decentralized Learning (2018, background): “s gradient information communicated by devices to the server. One line of work studies malicious devices who exploit anonymity and secure aggregation protocols to mount poisoning and backdoor attacks [7,12,82] on the system. Orthogonal to this line of work, are attacks [56,59] where the server is mo”
Neural Attention Distillation: Erasing Backdoor Triggers from Deep Neural Networks (2021, background): “Recent studies have shown that a backdoor attack can be conducted even without access to the training data (Liu et al., 2018b) or in federated learning (Xie et al., 2019; Bagdasaryan et al., 2020; Lyu et al., 2020).”
Reflection Backdoor: A Natural Backdoor Attack on Deep Neural Networks (2020, methodology): “ifferent to these methods, in this paper, we propose a natural reﬂection backdoor, which is stealthy, effective and hard to erase. Backdoor attacks have also been found possible in federated learning [2,48,60] and graph neural networks (GNNs) [65]. Latent backdoor patterns and properties of backdoor”
A Robust Privacy-Preserving Federated Learning Model Against Model Poisoning Attacks (2024, background): “These attacks have impeded the advancement of PPFL [28], [29], introducing encrypted poisonous updates rather than poisonous samples, making them more clandestine and potent [28], [29].”
A Survey on Federated Learning Systems: Vision, Hype and Reality for Data Privacy and Protection (2019, methodology): “[187] propose a distributed backdoor attack on FL.”
Privacy and Robustness in Federated Learning: Attacks and Defenses (2020, background): “Another weak but realistic attack scenario is backdoor poisoning [24], [25], [64].”
Anti-Backdoor Learning: Training Clean Models on Poisoned Data (2021, background): “Backdoor attacks on federate learning (FL) follow a different setting thus require different defense strategies [44, 45].”
FLDetector: Defending Federated Learning Against Model Poisoning Attacks via Detecting Malicious Clients (2022, methodology): “We choose this fraction because in the Distributed Backdoor Attack (DBA), the trigger pattern need to be equally splitted into four parts and embedded into the local training data of four malicious clients groups.”
Back to the Drawing Board: A Critical Evaluation of Poisoning Attacks on Production Federated Learning (2021, background): “Artificial backdoor attacks [67] aim to misclassify any input containing a backdoor trigger, hence these attacks are “indiscriminate” attacks.”