Federated Learning enables large-scale collaborative training across distributed devices. However, in massive-scale Internet-of-Things (IoT) deployments, ensuring the trustworthy sensor-level operations remains a critical challenge. We introduce a hierarchical framework that combines a three-tier architecture (devices → gateways → server) with a high-speed recursive proof system to enforce scalable zero-knowledge proofs (ZKPs). At the device level, each proof serves as a unified cryptographic commitment, binding the device’s identity, local data integrity, and training correctness into a single attestation. These proofs are then individually verified at intermediate gateways, and compressed into a single, succinct proof using a folding scheme inspired by Nova [1] - a state-of-the-art system that can excel at this task at best. The server then verifies a small number of batched proofs before aggregation, reducing workload (∼ 571× in data load) by replacing hundreds of thousands of individual proof and model update transmissions with just one per gateway. Our fully implemented R1CS precursor demonstrates resilience against various vectors (e.g., backdoor-style attacks,) achieves a ∼ 34× verification speedup on a 105-device network, and maintains both strong security and model performance. Our prototype, evaluated on an Internet-of-Vehicles (IoV) use case, demonstrates that recursive proofs add succinct overhead while providing scalable, robust integrity guarantees against adversarial environments.

Federated Learning (FL) enables multiple participants to collaboratively train a shared model without sharing their private data. However, FL remains vulnerable to malicious clients submitting incorrect updates to disrupt training. To address this, we formalize each client’s local training step as a Nondeterministic Polynomial-time (NP) statement, verifiable via zero-knowledge proofs (ZKPs) at every round. We propose Jolt-FL, the first general-purpose verifiable FL framework that immediately detects and excludes malicious clients upon their first dishonest action – without relying on heuristics, statistical assumptions, or multi-round analysis. Built on Jolt’s zkVM, a state-of-the-art zero-knowledge virtual machine (zkVM) developed by a16zcrypto, Jolt-FL guarantees training integrity and data privacy without trusted hardware or third-party intermediaries. By witnessing every computation step, it defends against a wide range of attack vectors, securely filtering dishonest updates even if up to 50% of clients are malicious, while preserving convergence and final model performance. To demonstrate feasibility, we implement a prototype featuring a complete end-to-end Convolutional Neural Network (CNN) for image classification using the MNIST dataset. To our knowledge, this is the first fully verifiable end-to-end CNN training under ZKPs without any custom circuit design. Our solution achieves competitive proof generation times, compact proof sizes, and low verification costs–all while preserving model accuracy on par with standard FL.

This paper investigates resource allocation for optical intelligent reflecting surfaces (OIRS) on high-altitude platforms (HAPs) supporting multiple UAV-mounted base stations. We propose a rate-optimized spatial resource allocation (R-SRA) scheme that first maximizes the number of QoS-guaranteed UAVs, then allocates remaining OIRS elements to boost the total system rate. Numerical results show R-SRA outperforms conventional methods, even with UAV mobility.

This paper proposes a joint design of probabilistic constellation shaping (PCS) and precoding to enhance the sum-rate performance of multi-user visible light communications (VLC) broadcast channels subject to signal amplitude constraint. In the proposed design, the transmission probabilities of bipolarM-pulse amplitude modulation (M-PAM) symbols for each user and the transmit precoding matrix are jointly optimized to improve the sum-rate performance. The joint design problem is shown to be a complex multivariate non-convex problem due to the non-convexity of the objective function. To tackle the original non-convex optimization problem, the firefly algorithm (FA), a nature-inspired heuristic optimization approach, is employed to solve a local optima. The FA-based approach, however, suffers from high computational complexity. Thus, using zero-forcing (ZF) precoding, we propose a low-complexity design, which is solved using an alternating optimization approach. Additionally, considering the channel uncertainty, a robust design based on the concept of end-to-end learning with autoencoder (AE) is also presented. Simulation results reveal that the proposed joint design with PCS significantly improves the sum-rate performance compared to the conventional design with uniform signaling. For instance, the joint design achieves$\mathbf {17.5\%}$and$\mathbf {19.2\%}$higher sum-rate for 8-PAM and 16-PAM, respectively, at 60 dB peak amplitude-to-noise ratio. Some insights into the optimal symbol distributions of the two joint design approaches are also provided. Furthermore, our results show the advantage of the proposed robust design over the non-robust one under uncertain channel conditions.

Uncrewed Aerial Vehicle-mounted Base Stations (UAV-BSs) have been envisioned as a promising solution to enable high-quality services in next-generation mobile networks. With inherent flexibility, one key challenge is placing the UAV-BSs adaptively to time-varying network conditions to maintain stable connections. Conventional methods mainly focus on optimizing UAV-BS deployment in static networks where users are static or with limited mobility. This study considers a dynamic network where multiple non-stationary UAV-BSs are deployed to serve mobile users with time-varying heterogeneous traffic. Incomplete downloads of users are backlogged in download queues at the UAV-BSs, and together with the newly requested traffic, the download queue size reflects the user’s instantaneous traffic demand. With constraints on the queue stability, we aim to maximize the user’s long-term mean opinion score (MOS), reflecting how the perceived data rate satisfies their traffic demand. Since the users’ location and traffic demand vary over time, we dynamically group users into clusters using a K-means-based algorithm and adjust the 3D location of UAV-BSs using an actor-critic deep reinforcement learning (DRL) framework. The actor module is encoded using a deep neural network (DNN) that obtains the UAV-BS’s current location and a traffic heatmap of users to predict the optimal movement for UAV-BSs. The critic module utilizes Lyapunov optimization to control the queue stability constraint and evaluate the actor’s decisions. Extensive simulations demonstrate the proposed method’s superior performance over conventional rate-maximization approaches.

This paper proposes a joint design of probabilistic constellation shaping (PCS) and precoding to enhance the sum-rate performance of multi-user visible light communications (VLC) broadcast channels subject to signal amplitude constraint. In the proposed design, the transmission probabilities of bipolar M-pulse amplitude modulation (M-PAM) symbols for each user and the transmit precoding matrix are jointly optimized to improve the sum-rate performance. The joint design problem is shown to be a complex non-convex problem due to the non-convexity of the objective function. To tackle the problem, the firefly algorithm (FA), a nature-inspired heuristic optimization approach, is employed to solve a local optima to the original non-convex optimization problem. The FA-based approach, however, suffers from high computational complexity. Therefore, we propose a low-complexity design based on zero-forcing (ZF) precoding, which is solved using an alternating optimization (AO) approach. Simulation results reveal that the proposed joint design with PCS significantly improves the sum-rate performance compared to the conventional design with uniform signaling. Some insights into the optimal symbol distributions of the two joint design approaches are also provided.

Federated learning (FL)-enabled digital twin (DT) has recently attracted research attention to bring intelligent applications. However, enabling the FL-enabled DT in vehicular networks becomes challenging due to vehicle mobility’s impact on communication channels. In this regard, we propose to deploy an unmanned aerial vehicle (UAV) as a relay node to support the vehicular network. The objective is to minimize energy consumption under the trade-off with the latency and accuracy constraints of the DT model via a joint optimization of local accuracy, the local computation frequency, relay decision, and transmission power. To do so, we derive instantaneous formulas to update the accuracy and latency constraints, then solve the proposed problem using an iterative algorithm with convex optimization techniques. Numerical results show that the proposed dynamic optimization for UAV-aided vehicular networks can reduce up to 39.9% of consumption energy compared to conventional methods.

This paper proposes a new implementation of the BBM92 protocol in satellite continuous-variable quantum key distribution (CV-QKD) systems using dual-threshold/direct detection (DT/DD). The proposed method is less complex and thus possibly cheaper than current discrete-variable (DV) and CV-QKD systems using coherent detection. We model and analyze the performance of the proposed system in the context that a satellite distributes secret keys to two legitimate users. The analytical results are derived by considering the channel loss, atmospheric turbulence-induced fading, and receiver noises. The Gaussian beam model is considered to evaluate the impact of geometrical spreading on the signal received by legitimate users and the possibility of being eavesdropped on. Based on the design criteria for the system and analytical results, we find suitable parameters for the transmitter and the receivers to properly achieve the QKD function for distributing secret keys between two legitimate parties.

This paper addresses the link-layer error-control solutions of high platform altitude (HAP)-aided relaying satellite free-space optical (FSO) systems for the Internet of Vehicles. Specifically, we propose a design of cooperative incremental redundancy (IR) hybrid automatic repeat request (HARQ) protocol to mitigate the transmission errors over the turbulence fading channels. The performance metrics, including average throughput and average frame delay, are analytically derived. The numerical results confirm the effectiveness of our proposed cooperative IR-HARQ protocol in FSO-based satellite-HAP-vehicle systems and support the proper selection of parameters. Monte-Carlo simulations are also performed to validate the correctness of theoretical results.

Satellite-based Internet access for the whole globe, a newly emerging market, has recently received much attention from both academia and industry. In this paper, we present an analytical investigation of transmission control protocol (TCP), which is the most popular protocol for various Internet applications, in free-space optical (FSO) communications based low earth orbit (LEO) satellite systems. Specifically, the throughput performance of the potential deployed TCP variants for high-speed and long-distance of FSO-based satellite networks, namely TCP Hybla and TCP Cubic, are analyzed. Additionally, the incremental redundancy hybrid automatic repeat request (IR-HARQ) protocol is employed to enhance the system performance over satellite-to-vehicles FSO links. The numerical results quantitatively demonstrate the impact of atmospheric turbulence on the TCP throughput and show that TCP Cubic outperforms Hybla in the low error-rate conditions while Hybla provides the better performance when the transmission errors happen more frequently. Monte Carlo simulations are also performed to validate the accuracy of theoretical derivations.

This paper studies the error-control protocol design for free-space optical (FSO) burst transmission in satellite communication systems. Specifically, we model and analyze the throughput performance of LEO-to-ground FSO systems over atmospheric turbulence channels when incremental redundancy hybrid automatic repeat request (IR-HARQ) protocols, which combine rate-compatible punctured convolutional (RCPC) code and sliding window ARQ, are employed. For this purpose, the time-varying behavior of atmospheric turbulence channels is first captured by a finite-state Markov chain. Then, the channel model is used to develop the burst loss model for IR-HARQ in order to analytically derive the system throughput. The results quantitatively show the impact of atmospheric turbulence on the throughput performance and support the optimal selection of system parameters. Monte Carlo simulations are also performed to validate the accuracy of theoretical derivations.