Federated learning (FL) enables collaborative model training without exposing local data, offering privacy benefits. However, its distributed nature makes it vulnerable to backdoor attacks, where adversaries manipulate training data or model updates to trigger attacker-chosen outputs. Existing defenses often fail under high proportions of malicious clients and struggle to balance robustness and model utility. This article proposes FedRAB, a collaborative FL defense framework using dynamic smoothing to mitigate backdoor threats. FedRAB remains effective even when over 50% of clients are malicious. In this framework, clients are categorized into three types: fully trusted clients, malicious but trusted clients and malicious and untrusted clients. The first two inject controlled perturbation noise into their local datasets, suppressing poisoning attacks while preserving accuracy. To address diverse and severe backdoor behaviors, the server applies dimensionality reduction followed by clustering to identify and filter out the most harmful malicious updates. This enhances both the accuracy and efficiency of malicious update detection. The server then clips and perturbs the remaining model updates, further strengthening defense against backdoors while preserving data diversity and generalization. We evaluate the effectiveness of FedRAB on various datasets. For example, on the MNIST dataset, when 65% of clients are malicious, FedRAB reduces the backdoor accuracy from 94.6% to 1.5%, while only decreasing the model's accuracy on benign samples by 1.2%.

Federated learning (FL) enables decentralized industrial-Internet-of-Things devices (also called clients) to share model parameters to build a joint model. Fair rewards, security of shared data, and transmission cost are the important factors that influence clients to participate in FL. Few existing works can solve these problems at the same time. Therefore, we propose a hierarchically fair and differentially private federated learning (HFDPFL), which regards the model itself as a reward to promote fairness. Reputation is used to measure the client's contribution to FL, and clients with high reputation will be rewarded with high accuracy models. In order to ensure the security of the shared data and reduce communication overhead, we implement differentially private gradient compression based on compressed sensing, which achieves differential privacy protection of gradients and improves communication efficiency. Extensive experiments are conducted to demonstrate the superiority of HFDPFL in terms of fairness, privacy preserving, and communication efficiency.