BackDFL is presented, a unified benchmark for systematically evaluating DFL under realistic and adaptive backdoor attacks, and demonstrates that both state-of-the-art Byzantine-robust DFL methods and adapted FL backdoor defenses fail under modest malicious participation rates, especially in heterogeneous settings.
Abstract
Decentralized Federated Learning (DFL) promises trust-free collaborative learning by replacing the centralized parameter server with peer-to-peer model exchange. However, this architectural shift fundamentally reshapes the threat landscape. Without globally coordinated aggregation, DFL becomes particularly susceptible to backdoor attacks, in which malicious participants implant persistent hidden behaviors while maintaining high clean-task performance. In this paper, we argue that the robustness of DFL has been significantly overestimated. Existing studies rely on simplified threat models, non-adaptive adversaries, fragmented evaluation protocols, inconsistent communication topologies, and ad hoc training configurations, leading to an incomplete understanding of DFL security. To address these limitations, we present BackDFL, a unified benchmark for systematically evaluating DFL under realistic and adaptive backdoor attacks. Through extensive experiments, BackDFL exposes critical failure modes of decentralized learning. Our results demonstrate that both state-of-the-art Byzantine-robust DFL methods and adapted FL backdoor defenses fail under modest malicious participation rates (as low as 15%), especially in heterogeneous settings, while their robustness varies substantially across communication graph topologies.
Federated Learning (FL) is vulnerable to backdoor attacks, where adversaries can stealthily manipulate the global model. Most existing defense methods are developed under IID assumptions, an assumption that rarely holds in practice. In highly non-IID settings, heterogeneous data distributions across clients make it difficult to distinguish malicious updates from benign ones, particularly when benign clients exhibit atypical patterns due to minority-class data. To address this challenge, existing defenses operate at different levels of granularity. Coarse-grained methods perform client-level filtering, which often mistakenly excludes benign clients under non-IID conditions. Fine-grained methods instead analyze data at the sample level for more precise detection, but typically rely on explicit per-sample gradient analysis, leading to substantial memory and computational overhead. As a result, defending against backdoor attacks in non-IID environments involves a fundamental trade-off between robustness and computational efficiency. To address this challenge, we propose FedRGD, a federated risk-guided dynamic defense framework that enables efficient fine-grained protection. FedRGD maps sample-level risks into structured parameter masking without requiring explicit per-sample gradient storage. It combines feature inconsistency detection with lightweight masking and robust aggregation to achieve both accuracy and efficiency. Extensive experiments on CIFAR-10 and Fashion-MNIST demonstrate that FedRGD consistently reduces the attack success rate while maintaining high main-task accuracy, achieving a favorable security-utility balance with low computational overhead.
Ruiying Wang· Poster Volume 0008 The 2026...· 0 citations
This work proposes CAEBA (Conditional AutoEncoder Backdoor Attack), a dynamic hidden backdoor framework that uses a conditional autoencoder to generate target-aware and visually stealthy triggers while progressively implanting the backdoor through federated optimization.
Federated learning is appealing for privacy-sensitive network systems, yet its practical deployment remains hindered by the following three recurring challenges: (1) client drift under non-IID data, (2) vulnerability to corrupted updates, and (3) the communication cost of repeated model exchange. Most existing approaches address these issues in isolation. While analytically convenient, this separation often fails to reflect real-world conditions. For instance, defenses against poisoning may suppress useful updates, while personalization and compression can alter the aggregation geometry itself. In this paper, we study these effects jointly and propose URP-FL, a compact training framework that integrates reliability-aware aggregation, local regularization for drift control, and sparse client uploads. We provide theoretical analysis establishing a convergence bound with distinct terms capturing optimization error, data heterogeneity, and adversarial impact. Experiments on a non-IID image classification benchmark with sign-flip and label-flip attacks demonstrate the benefits of the unified design. Compared to FedAvg and FedProx, this URP-FL maintains accuracy under attack while reducing transmitted parameters by approximately 75%. Rather than presenting a production ready system, it offers a reproducible and technically coherent step toward federated learning that is more robust under realistic conditions.
Hua Kun, Wei Wang· 2026 International Conferenc...· 0 citations
Federated Learning (FL) inherently preserves privacy but remains highly vulnerable to backdoor attacks due to its open participation architecture. Existing defenses face two fundamental limitations: first, screening-based aggregation strategies prove ineffective against advanced cross-round attacks where adversaries progressively poison model parameters through multi-round collaboration; second, mitigation techniques often cause significant accuracy degradation due to the deep entanglement between backdoor and primary task parameters. To address these challenges, we propose Fed-CBE, a novel client-side defense algorithm that eliminates backdoors through three synergistic mechanisms: 1) periodic alternating layer resetting disrupts deep parameters to dismantle cross-round backdoor accumulation; 2) indiscriminate forgetting employs entropy maximization on non-ground-truth classes to decouple backdoor associations without prior trigger knowledge; and 3) knowledge distillation with historical local models restores primary task performance. Extensive evaluations on three benchmark datasets and model architectures demonstrate that Fed-CBE achieves highly competitive robustness, limiting attack success rates to near-zero levels in most settings and keeping them exceptionally low even under high malicious-client ratios without compromising primary task performance, significantly outperforming existing defenses.
Chunhai Li, Yun-Hui Shen, Ming Xie et al.· IEEE Transactions on Informa...· 0 citations
The rapid adoption of Internet of Things (IoT) devices has accelerated the need for privacy-preserving machine learning techniques, such as federated learning (FL). However, the decentralized and collaborative nature of FL makes it vulnerable to backdoor attacks, where adversaries locally update their malicious models before contributing to the global aggregation, subtly injecting backdoors without degrading the normal performance. An affected model behaves as expected during regular operations but exhibits malicious behavior when an embedded trigger is presented. In this article, we propose a novel self-supervised contrastive-learning-based approach to detect and mitigate backdoor attacks in FL within IoT environments. Unlike conventional reverse-engineering methods that iterate through each class in the dataset to reconstruct triggers, our approach directly regenerates triggers from compromised global models without class iteration. This is achieved by comparing last-layer feature representations of a potentially compromised model with those of a relatively clean model under the guidance of contrastive loss. The reverse-engineered trigger is then leveraged to patch the global model and remove the backdoors. We evaluate our method on three benchmark datasets under two federated backdoor attack scenarios, simulating IoT device collaborations. Extended experiments are also conducted on a transformer-based model and two mitigation methods to assess the robustness of our approach. Our results demonstrate that while traditional reverse-engineering techniques are effective in centralized settings, they struggle to detect backdoors in FL. Comparatively, our method is resilient against backdoor attacks across various settings. In addition, our method is more time-efficient because of its capability of generating the backdoor trigger directly without iterating through all classes.
Hal Ferguson, Rui Ning, Hongyi Wu et al.· IEEE Internet of Things Jour...· 0 citations
Federated Learning (FL) is vulnerable to backdoor attacks because of its distributed nature in edge computing scenarios. Existing defense methods show limited efficacy as they overlook the deviations among benign local updates caused by statistical heterogeneity and the stealthiness of backdoor attacks. To tackle these issues, we propose FedDAB, a two-phase method that combines local contrastive regularization with alignment checking, to defend against backdoor attacks. In the first phase, FedDAB introduces a novel model-contrastive term into the local objective to enhance direction and magnitude consistency among benign updates. In the second phase, FedDAB employs an alignment checking strategy to evaluate each local update in terms of overall-direction alignment and parameter-level alignment with historical information, excluding updates that exhibit abnormal alignment patterns from global aggregation. We theoretically prove FedDAB's robustness with a convergence rate of $\mathcal{O}(1/T)$. Extensive experiments show that FedDAB outperforms existing defense methods against backdoor attacks.
Hongliang Zhang, Zhongyuan Yu, Guijuan Wang et al.· 0 citations