Federated learning (FL), as a distributed machine learning paradigm with privacy protection, has garnered significant attention since it prevents the exchange of raw local data. However, FL remains vulnerable to poisoning attacks, including data contamination and gradient manipulation. Moreover, attackers may launch individual or collusive attacks, complicating the identification of malicious clients. To address these challenges, we propose a universal poisoning defense framework incorporating three key strategies. First, we decouple client identities from gradients through anonymous obfuscation and enhance privacy with differential noise injection. Second, we detect potential detect potential collusive attackers via a joint similarity-based approach. Third, we apply an iterative low rank approximation-based anomaly detection to amplify discrepancies between benign and malicious clients and progressively filter out attackers. We theoretically demonstrate that anonymous obfuscation can enhance the privacy protection capability of differential privacy. Additionally, experimental results further validate that our scheme is comparable to or outperforms state-of-the-art defense methods against a variety of data and model poisoning attacks.

For current and future wireless communication networks, accurate Channel State Information (CSI) is pivotal for delivering high-quality services for countless devices. Traditional Channel Estimation (CE) methods grapple with channel aging, leading to inaccuracies. In contrast, we obtain the precise CSI by predicting channels based on neural networks, which have outstanding performance in feature extraction, data generation and prediction. Centralized Learning (CL) methods impose high demands on computational resources, storage and communication overhead. Moreover, it poses a risk to user data privacy. Consequently, we propose a Federated Learning-based Long Short Term Memory (FL-LSTM) Neural Network for real-time CSI prediction. We validate our model on the simulation dataset and measurement datasets. Simulation results show that our approach can predict the CSI data effectively, and its prediction performance approximates to Centralized Learning-based Long Short Term Memory (CL-LSTM) Neural Network. In addition, the global model is capable of generalization, which can be transferred to a new base station (BS) to accelerate the convergence of the local model and reduce the prediction error.