Federated Learning (FL) is a machine learning training method that leverages local model gradients instead of accessing private data from individual clients, ensuring privacy. However, the practical implementation of FL faces significant challenges. Heterogeneous clients and edge devices with varying computational abilities and unreliable communication channels introduce latency issues to the algorithm. Furthermore, the algorithm is susceptible to attacks from malicious clients, allowing them to insert unwanted updates while benefiting from the global model. These challenges severely impact the algorithm’s performance, rendering it unsuitable for real-time applications. To address these issues, we propose FedHSP, a comprehensive system that tackles device heterogeneity and protects against various forms of attacks. FedHSP incorporates multiple model complexities to accommodate heterogeneous clients. Additionally, it employs a Variational Auto Encoder with dynamic thresholding to detect and eliminate malicious clients. In this paper, we demonstrate FedHSP’s effectiveness in detecting model poisoning attacks. We also show the mitigation of malicious model updates sent to the server. The evaluation is done using the MNIST dataset under various settings. Our experiment results show the drastic performance deviation due to attacks and the successful detection and mitigation with the proposed system.

We consider the problem of data poisoning attack detection in a federated learning (FL) setup with differential privacy (DP). Local DP in FL ensures that privacy leakage caused by shared gradients is controlled by adding randomness to the process. We are interested in studying the effect of the Gaussian mechanism in the detection of different data poisoning attacks. As the additive noise from DP could hide poisonous data, the effectiveness of detection algorithms should be analyzed. We present two poisonous data detection algorithms and one malicious client identification algorithm. For the latter, we show that the effect of DP noise decreases as the size of the neural network increases. We further demonstrate this effect alongside the performance of these algorithms on three publicly available datasets.

The integration of Artificial Intelligence and Machine Learning (AI/ML) into 5G and forthcoming 6G networks critically relies upon User Equipment (UE) collected data. However, this dependency creates a vulnerability to data poisoning, as current defenses on UEs lack contextual understanding of the collected data’s intended use by data consumers. This paper proposes SANITGUARD-UE, a novel data consumer purpose-driven solution designed to prevent the ingress of poisoned data. It seamlessly integrates with the 3rd Generation Partnership Project (3GPP) architecture by enabling UE sanitization capability signaling, Data Collection Configuration Function (DCCF)-configured purpose-driven data sanitization and an enhanced L2 header. Leveraging the use case of cell load estimation, empirical experiments demonstrated that SANITGUARD-UE significantly improves AI/ML model accuracy, outperforming purpose-agnostic sanitization by a factor of 3.73 to 7.66. This research establishes the essential role for purpose-driven data sanitization to build the trustworthiness and to optimize the performance of AI/ML services in next-generation mobile networks.