Abstract The proliferation of Cyber-Physical Systems (CPS) across critical infrastructure has created an unprecedented attack surface where digital threats may precipitate catastrophic physical consequences. As conventional centralized security paradigms fail to address the scale and complexity of these environments, Federated Learning (FL) has emerged as a transformative approach, enabling collaborative, edge-native anomaly detection without centralizing sensitive data. This paper presents a comprehensive survey and critical analysis of the state-of-the-art in securing CPS through advanced FL. We introduce a novel multi-axis taxonomy that systematically categorizes the field by architecture, detection methodology, application domain, and privacy-preservation scheme. Building on this analysis, we synthesize these findings into a prescriptive framework to guide the selection of appropriate security archetypes for different CPS domains. Through this lens, we deconstruct the—the trade-off between accuracy, communication, and privacy- that governs every FL design. Our analysis synthesizes the dominant trends, including the convergence of deep learning with edge computing and the increasing sophistication of privacy-enhancing technologies. We further identify critical research gaps, including the scarcity of physical testbeds, limited resilience against advanced adversarial attacks, and underdeveloped explainability. The paper concludes by defining the critical frontiers for future research, emphasizing the need to resolve the inherent tension between FL’s privacy goals and the transparency requirements of Explainable AI (XAI) to build truly trustworthy systems.

One-dimensional signal decomposition is a well-established and widely used technique across various scientific fields. It serves as a highly valuable pre-processing step for data analysis. While traditional decomposition techniques often rely on mathematical models, recent research suggests that applying the latest deep learning models to this very ill-posed inverse problem represents an exciting, unexplored area with promising potential. This work presents a novel method for the additive decomposition of one-dimensional signals. We leverage the Transformer architecture to decompose signals into their constituent components: piecewise constant, smooth (trend), highly-oscillatory, and noise components. Our model, trained on synthetic data, achieves excellent accuracy in modeling and decomposing input signals from the same distribution, as demonstrated by the experimental results. • We study additive decomposition of 1D signals with Transformers. • We define a neural architecture for the problem, based on the Transformer encoder. • Our method is more effective and orders of magnitude faster than variational ones. • The proposed method automatically detects the absence of a component.

Federated Learning (FL) enables collaborative model training without sharing raw data, making it attractive for privacy-preserving applications at the wireless edge. However, when executed over real 5G networks, FL performance degrades due to uplink congestion, heterogeneous client capabilities, and intermittent connectivity. Most existing approaches attempt to mitigate these issues indirectly by optimizing clients (through adaptive participation, local training, or selection strategies) or by optimizing models (via pruning, quantization, or compression), but they ignore potential network bottlenecks. This paper introduces FLAG, an FL architecture that embeds innetwork aggregation directly into 5G gNodeBs, transforming the network into an active participant in the learning process. In particular, FLAG performs parameter aggregation at line rate within the 5G Service Data Adaptation Protocol layer and incorporates three mechanisms: Partial-Contribution Correction for loss-tolerant averaging, a timer-driven pipeline for real-time scheduling, and a deadline-based grouping strategy to mitigate stragglers. Experiments with realistic wireless emulation show that FLAG achieves up to 5.1× faster time-to-accuracy and maintains accuracy within 0.8% of a loss-free baseline, while reducing gNB-to-server bandwidth by aggregating pergNB rather than per-client. FLAG requires no modifications to clients or the parameter server, demonstrating how 5G-aware system design can make federated learning scalable, efficient, and resilient under real-world wireless conditions.

While previous optimization results have suggested that deep neural networks tend to favour low-rank weight matrices, the implications of this inductive bias on generalization bounds remain underexplored. In this paper, we apply a chain rule for Gaussian complexity (Maurer, 2016a) to analyze how low-rank layers in deep networks can prevent the accumulation of rank and dimensionality factors that typically multiply across layers. This approach yields generalization bounds for rank and spectral norm constrained networks. We compare our results to prior generalization bounds for deep networks, highlighting how deep networks with low-rank layers can achieve better generalization than those with full-rank layers. Additionally, we discuss how this framework provides new perspectives on the generalization capabilities of deep networks exhibiting neural collapse.

In recent decades, applications in environmental sustainability, education, and housekeeping have become increasingly distributed and sophisticated, leveraging a wide range of devices to perform complex tasks. While a large number of agents can reduce computation time, managing these distributed systems presents significant challenges due to resource constraints such as power consumption and storage. To address this, the literature has explored various model compression techniques, such as pruning, to optimize performance in distributed environments. In this paper, we propose DFL-Trim, a solution for trimming models in Decentralized Federated Learning (FL) that meets network constraints while maintaining satisfactory performance. We demonstrate how pruning can be implemented in decentralized settings, analyze its effect on bandwidth usage, and discuss the trade-offs between compression and model accuracy.

De-authentication Denial of Service (DoS) attacks in wireless networks allow adversaries to maliciously disassociate devices, interrupting communication and effectively denying service. The 802.11w protocol was designed to counter this issue using Protected Management Frames (PMF). However, our analysis reveals that during de-authentication DoS attacks, throughput drops significantly, and client disconnections may occur, exposing the limitations of the 802.11w protocol. Extended Berkeley Packet Filter (eBPF) and eXpress Data Path (XDP) technologies, recently adopted in wired networks to enhance packet processing efficiency, remain largely unexplored in wireless networks and their unique challenges, such as those posed by the 802.11 protocol. In this paper, we introduce a novel approach that integrates eBPF/XDP into the mac80211 Linux kernel module to mitigate de-authentication attacks in near real-time with minimal overhead. Our solution partially overcomes the shortcomings of the 802.11w protocol, offering a more robust defense.

In the context of NextG Wireless Networks, addressing the challenges of wireless communication link reliability is paramount to ensure efficient Distributed Learning systems. However, many recent solutions have overlooked key challenges, such as packet-level losses and the impact of TCP retransmissions, which are crucial for the robustness of these systems. In this paper, we propose the integration of fountain codes into the distributed learning process to offer a robust mechanism to counteract packet loss. Specifically, we propose a cumulative strategy logic based on fountain codes specifically tailored for packet exchanges in Distributed Learning applications. Our evaluation shows that fountain codes significantly enhance the efficiency and reliability of distributed learning model updates under severe packet loss conditions, e.g., a packet reduction of ≈ 84% (≈ 60%) at the UE (gNB) side compared to traditional TCP methods when packet loss probability reaches 0.9 in Federated Learning context. However, under low packet loss scenarios, fountain codes computational overhead becomes non-negligible. These results highlight the potential of fountain codes to serve as a robust alternative to conventional communication protocols in distributed learning systems, particularly in environments characterized by unstable network conditions.

Network slicing in next-generation wireless networks is a mechanism that ensures isolation and optimal resource distribution among users under constraints imposed by limited information availability at Base Stations (BS). While several strategies to optimize wireless slices have been proposed, this poster introduces an approach to resource allocation in 5G networks that integrates detailed end-to-end flow-level data at the transport layer to refine the NextG resource block allocation process. In particular, we present an architecture that leverages the host’s network stack and a novel in-network processing and scheduling component to optimize the dynamic resource allocation process in slicing. The proposed system design, illustrated through a comprehensive system architecture, aims to distribute Resource Blocks (RBs) effectively, accounting for the unique transport-level metrics of each flow. We present some initial evaluation results with an event-driven simulator, reflecting diverse traffic types and leveraging the Rayleigh fading channel model to simulate dynamic user movement. Our findings demonstrate a promising improvement in flow completion times and a reduction in packet loss, compared to traditional allocation methods such as Round Robin and Proportional Fair schemes, typically deployed in 5G production networks.

Every day, large amounts of sensitive data are distributed across mobile phones, wearable devices, and other sensors. Traditionally, these enormous datasets have been processed on a single system, with complex models being trained to make valuable predictions. Distributed machine learning techniques such as Federated and Split Learning have recently been developed to protect user data and privacy better while ensuring high performance. Both of these distributed learning architectures have advantages and disadvantages. In this article, we examine these tradeoffs and suggest a new hybrid Federated Split Learning architecture that combines the efficiency and privacy benefits of both. Our evaluation demonstrates how our hybrid Federated Split Learning approach can lower the amount of processing power required by each client running a distributed learning system, and reduce training and inference time while keeping a similar accuracy. We also discuss the resiliency of our approach to deep learning privacy inference attacks and compare our solution to other recently proposed benchmarks.

Localization has achieved great attention in 5G networks, pushed by standardization. However, experimentation in 5G networks lacks the integration of network function modules designed for localization. We present our implementation of the 5G Localization Management Function. It complies with the 3GPP standard and OpenAirInterface, the most advanced framework that implements a full 5G-New Radio stack. We show that we are able to extend the functionality of OpenAirInterface, enabling location services. Finally, we demonstrate that the tool's performance satisfies the 5G Key Performance Indicators required by 3GPP for localization.