Aug 2026· Tsinghua Science and Technology· 0 citations
TL;DR
This work proposes HFL-Lite, a hierarchical federated learning framework that achieves practical privacy without cryptographic primitives, and shows that HFL-Lite reduces per-round communication to 0.5 KB and sensor-side latency to 9 ms while delivering competitive accuracy.
Abstract
The rapid growth of sensor-equipped edge devices calls for machine learning paradigms that are both lightweight and privacy-preserving. Federated learning (FL) enables distributed AI on edge data but still demands heavy forward and backward propagation on resource-constrained sensors. Split learning (SL) offloads computation to a server, yet its intermediate activations can leak private information. Homomorphic encryption closes this gap at the cost of prohibitive computation and bloated communication. We propose HFL-Lite, a hierarchical federated learning framework that achieves practical privacy without cryptographic primitives. Each sensor encodes its raw data into a compact sparse feature code via a lightweight sensor encoder. Only this code, which is insufficient to reconstruct the original data, leaves the device. An edge server synthesizes surrogate training samples from the uploaded codes, trains a teacher model, and distills knowledge back to the sensors. Because the sensors perform only one-way uploads, backward propagation, gradient transmission, and plaintext activation leakage are all eliminated. Experiments on the UCI HAR dataset show that HFL-Lite reduces per-round communication to 0.5 KB and sensor-side latency to 9 ms while delivering competitive accuracy.
Mobile Edge Computing (MEC) enables resource-constrained Internet of Things (IoT) devices to offload computation-intensive workloads to nearby edge servers, reducing latency and energy consumption. However, wireless offloading exposes transmitted data to eavesdropping, raising serious privacy concerns. Existing physical-layer security techniques either introduce additional overhead or fail to conceal semantic information. This paper proposes a diffusion-based semantic encoding framework for secure and efficient computation offloading in MEC systems. By applying a forward diffusion process, task inputs are transformed into approximately noise-like latent representations whose distribution is statistically close to a standard Gaussian distribution prior to transmission. Authorized edge servers equipped with learned reverse diffusion models can reliably recover task-relevant semantics, while intercepted representations reveal negligible information. We analyze the secrecy and robustness of diffusion-based encoding from an information-theoretic perspective and integrate the proposed encoder-decoder into a practical MEC offloading pipeline. Extensive experiments under realistic wireless conditions demonstrate that diffusion-based encoding significantly improves secrecy, reduces transmission overhead, and maintains high task performance. Compared with autoencoder and variational autoencoder baselines, the proposed approach offers substantially stronger resistance to reconstruction and inference attacks while preserving computational efficiency at the edge.
Nowadays, split federated learning (SFL) has emerged as an effective paradigm for enabling privacy-preserving collaborative intelligence across heterogeneous devices with limited computation. However, SFL incurs significant communication overhead in wireless networks due to the uplink transmission of high-dimensional smashed data, which degrades network efficiency. To mitigate the communication bottleneck, we propose a prototype-based SFL framework ProtoSFL. Specifically, each selected client computes local prototypes for observed classes and uploads them to the server. Based on the received prototypes, the server derives global prototypes and optimizes a weighted objective that combines classification loss with prototype alignment loss. The server then updates the model accordingly and returns personalized prototype gradients to the clients. Simulation results verify the effectiveness of ProtoSFL in reducing communication overhead, achieving a substantial reduction in uplink communication, while maintaining competitive testing accuracy under various heterogeneous data settings compared with SFL baselines.
Personalized federated learning (PFL) is a promising paradigm for collaborative learning over distributed devices, where edge nodes collaboratively train personalized models without sharing raw data. Although PFL addresses data heterogeneity by learning client-specific models, it still suffers from substantial uplink and downlink communication costs when exchanging high-dimensional parameters in bandwidth-constrained systems. Recent one-bit methods achieve extreme compression, but they usually rely on a single thresholding rule applied to the whole model. This design has two limitations. First, it overlooks layer-wise differences in parameter distributions and quantization sensitivities. Second, a single threshold provides only coarse binary information and cannot capture fine-grained variations in parameter distributions. To address these issues, we propose a communication-efficient PFL framework via layer-wise multi-threshold random sketching. In the proposed method, each layer is assigned its own set of quantization thresholds, so that the compressed representation can adapt to layer-specific statistics while using multiple intervals to provide a finer low-bit description of sketched parameters. The proposed method supports bidirectional communication using compact low-bit sketches and improves the communication-accuracy tradeoff compared with existing one-bit compression approaches.
Xu Zhang, Xing-Yu Hou, Jia-Cheng Cheng et al.· 0 citations
Federated fine-tuning is bottlenecked by communication: FedAvg and pseudo-gradient schemes transmit a payload that scales with the model, and gradient compression shrinks it by only a constant factor. We take a different lever. Mapping networks generate a network's weights from a small trainable latent through a frozen affine projection; because the map is shared and affine, averaging latents is exactly averaging the generated weights. We turn this into a practical low-bandwidth federated channel with two changes: a low-rank, seed-regenerable factorisation of the projection (cutting generator memory from ~80 GB to ~10 MB), and a delta formulation $\theta = \theta^{\mathrm{pre}} + U V^{\top} z$ that learns an additive correction around a shared centrally-pretrained base -- federated fine-tuning, which is what makes the method work at scale. A frozen orthogonal classifier head further removes the head from the payload while improving accuracy. On CIFAR-100 with ResNet-18+GroupNorm, our method (FLITE, Federated Low-rank Iterative Training Engine) communicates 1,280 floats (~5 KB) per client per round -- an 8718x reduction -- and reaches 74.67%, within ~0.5 pp of full-weight FedAvg. The averaging identity holds to floating-point precision ($6 \times 10^{-8}$); the method sits one to two orders of magnitude below PowerSGD and top-k on the bandwidth-accuracy Pareto; it matches or exceeds full-weight FedAvg under strong non-IID skew. int4 latents reach 648 bytes per round at unchanged accuracy, whereas int4 full-weight FedAvg collapses to chance.
Distributed Learning (DL) is a fundamental paradigm for large-scale model training in mobile and edge computing. However, its practical adoption is often plagued by performance degradation due to straggler and Byzantine nodes, compromising overall robustness and efficiency. Although coded computing provides theoretical solutions to these challenges, existing implementations are constrained by two critical limitations: inefficient decoding and expensive verification. To address these two issues, we propose FVCC, a fast and verifiable coded computation framework that enables robust DL. Specifically, we employ two-dimensional Shift-and-Add (SA) encoding and ZigZag Decoding (ZD) strategies for large-scale matrix-matrix multiplications prevalent in DL. To improve decoding efficiency, we propose a bidirectional two-dimensional ZD (4D-ZD) algorithm that enables parallel processing, significantly reducing recovery latency. Moreover, we introduce a lightweight verification mechanism based on Freivalds’ algorithm to defend against Byzantine attacks with low overhead. Finally, we conduct a comprehensive theoretical analysis and experimental evaluation. Empirical results demonstrate that 4D-ZD achieves approximately $2\times $ faster decoding compared to the state-of-the-art scheme. Moreover, FVCC maintains model accuracy while reducing the training time by approximately 38.55% for small-scale and 42.87% for large-scale DL tasks compared to state-of-the-art baselines.
Lingling Wang, Linqing Xu, Yufei Liu et al.· IEEE Transactions on Informa...· 0 citations
This work proposes a lightweight and information-theoretically secure aggregation framework that securely computes the majority vote (MV) polynomial through single-round secure multiplication, ensuring end-to-end information-theoretic security under the honest-majority assumption while revealing only the final aggregated sign to the server.
