Generative diffusion models deliver remarkable synthesis quality but remain impractical for resource-limited Internet of Things (IoT) devices due to their substantial computational and memory demands. To bridge this critical gap, we present a comprehensive, multi-stage optimization framework that systematically reduces model size while meticulously preserving generative fidelity. The framework integrates an efficient backbone architecture designed for inherent lightness, a sensitivity-guided fine-grained pruning strategy that strategically removes redundant parameters to achieve high sparsity, and a novel distribution-aware quantization algorithm based on Gaussian Mixture Models (GMMs) to compress weights and activations with minimal quality degradation. Extensive validation across multiple diffusion architectures (DDPM, DDIM, SGM) and diverse datasets demonstrates the framework’s strong generalizability, achieving up to 79% model sparsity while preserving generative fidelity. To showcase practical utility, we demonstrate that our framework produces a compressed model compatible with standard mobile deployment toolchains, realizing a significant reduction in the on-device memory footprint required for inference. This work offers a robust and generalizable methodology for enabling advanced generative AI on a wide spectrum of edge and IoT platforms. Code is available at: https://github.com/mitchell-cheng/compress_diffusion.

Neighbor discovery poses significant challenges in Underwater Optical Wireless Sensor Networks (UOWSNs) due to the unique characteristics of directional transceivers, line-of-sight communication, and mobility induced by water currents. Traditional methods typically rely on prerequisites and prior knowledge, such as centralized coordination, time synchronization, and information about the number of neighbors, which are often unavailable or impractical in underwater environments. In this paper, we make the first attempt to address the issue ofRobust andEfficientNeighborDiscovery (termed the REND problem) in UOWSNs with zero-knowledge. Here, zero-knowledge refers to the capability that enables sensors to identify neighbors in dynamic underwater optical channel conditions without prerequisites or prior knowledge. We design a zero-knowledge distributed directional neighbor discovery scheme inspired by gear meshing. We then propose a deterministic algorithm for the REND problem based on theoretical analysis. Additionally, to further reduce the discovery delay for the periodic REND problem, we develop a greedy-based approximation algorithm with a performance guarantee. Finally, extensive simulations demonstrate that the proposed scheme reduces the discovery delay by 34.9% on average and achieves an additional 54.4% reduction for periodic neighbor discovery. Furthermore, test-bed experiments are carried out to verify the applicability of our zero-knowledge scheme in real-world scenarios.

WiFi-based identity recognition differs from traditional identification technologies as it is not limited by lighting conditions and does not require dense, specialized sensors or wearable devices. This makes it valuable in modern human–machine interactions. However, the diversity of real-world environmental conditions substantially limits the application of existing WiFi-based identity recognition algorithms, particularly when applied across different environments. As a solution, we introduce the incremental wavelet capsule (IWC) model, which combines a newly designed wavelet convolution layer with a capsule network to accelerate precise feature extraction. We adopt a hybrid incremental learning strategy, solving the catastrophic forgetting1problem in cross-environment tasks and enabling the model to adapt to new environments in the data stream without forgetting the original environment. Furthermore, we developed a customized data augmentation method for WiFi signals, enhancing the model’s adaptability and stability across various environments. Experimental results show that the IWC model achieves an average recognition accuracy of 97.36% across five different environments and maintains an accuracy of 91.5% even when only 5% of the training data from a new environment is used. These findings demonstrate the model’s robust performance and practicality in cross-environment scenarios.

Recently, for indoor personal drones, there have been a number of challenges in localizing users, such as how to accurately identify their location and identification. This paper proposes a distance measurement method for indoor Unmanned Aerial Vehicle (UAV) localization combining fine-time measurement (FTM) ranging equipment and the Deep Neural Network (DNN) model. Specifically, we can calculate the distance between the UAV and user by measuring the signal round-trip time. However, the indoor ranging accuracy of the FTM protocol is inevitably affected by the multipath and non-line-of-sight (NLOS). Besides, we prove that the FTM ranging under multipath is related to either the distance between the transceiver pair or the length of the reflected path relative to the direct one. Moreover, we also study the relationship between the FTM ranging error and the response of the multipath channel. Furthermore, we design an FTM error calibration model based on physical layer (PHY) information by using the DNN model, termed DeepF, which can not only automatically distinguish environmental characteristics but also estimate the length of propagation paths of a signal. The designed DeepF can adopt the DNN model to extract the time domain information of channel state information (CSI) and learn the nonlinear mapping between delay, power, and mean error from signal features. Finally, we use a trained model to calibrate the FTM error and predict the user location. Experimental results show DeepF significantly improves the ability of indoor UAVs to localize users.

Gait is regarded as a unique feature for identifying people, and gait recognition is the basis of various customized services of the IoT. Unlike traditional techniques for identifying people, the Wi-Fi-based technique is unconstrained by illumination conditions and such that it eliminates the need for dense, specialized sensors and wearable devices. Although deep learning-based sensing models are conducive to the development of Wi-Fi-based identification, the latter technique relies on a large amount of data and requires a long training time, where this limits the scope of its use for identifying people. In this study, we propose a Wi-Fi sensing model called Wave-CapNet for human identification. We use data processing to eliminate errors in the raw data so that the model can extract the characteristics in channel state information (CSI). We also design a dedicated adaptive wavelet neural network to extract representative features from Wi-Fi signals with only a few epochs of training and a small number of parameters. Experiments show that it can identify human gait with an average accuracy of 99%. Moreover, it can achieve an average accuracy of 95% by using only 10% of the data and fewer than five epochs and outperforms state-of-the-art (SOTA) methods.

Underwater Optical Wireless Sensor Networks (UOWSNs) are gaining an increasing demand in industrial and commercial applications as they can achieve high-speed communication. However, prior arts concentrate on promoting the performance of UOWSNs, while the reliability issue has not been fully addressed. In this paper, we propose a novel reliable data delivery scheme based on a cluster structure. First, we determine the orientation of each sensor for directional optical communication, which aims to establish reliable next-hop links among sensors. We formalize such an orientation problem into a submodular function maximization problem and propose a greedy method with an approximation ratio guarantee to solve it. Then, a cluster head designation scheme is developed to improve the data delivery success rate while minimizing the number of cluster heads. Finally, extensive simulations are conducted to demonstrate the effectiveness of the proposed scheme. The results reveal that compared with other algorithms, the proposed scheme can ensure a data delivery success rate of over 98.5 % while only keeping 45.3% fewer cluster heads. Furthermore, test-bed experiments are carried out to verify the applicability of the proposed scheme in practical applications.

Underwater Optical Wireless Communication (UOWC) enables high-speed data transmission among Autonomous Underwater Vehicles (AUVs). However, due to cost and weight constraints, AUVs can only carry a limited number of directional optical transceivers. This implies that each AUV can communicate with only 1 to 2 neighbors simultaneously, complicating the establishment of an Underwater Optical Wireless Communication Network (UOWN). To address the networking problem with the degree constraint, we propose a topology generation method based on Hamiltonian paths. The topology achieves improved global connectivity at the cost of local optimality while satisfying the communication device limitations of AUVs. Preliminary results show that the generated topology can reduce the average communication overhead.

The ability to detect which wireless devices are belonging to the same person from Wi-Fi access point (AP) enables many potential Internet-of-Things (IoT) applications, including continuous authentication and user-oriented devices isolation. The existing cryptographic-based solutions are not suitable for IoT devices with limited power and computing capabilities. The development of electronics and chip technology makes it possible to deploy machine learning (ML) algorithms on APs. In this article, we propose an on-body device clustering (OBDC) scheme. First, the OBDC extracts the trajectory and gait patterns from wireless signals when the user is moving. Second, it utilizes a hierarchical clustering algorithm to measure the similarity of wireless signal patterns between devices. Finally, if the devices are clustered into the same cluster, they are considered to be carried by the same person. Our real-world experimental results show that the devices from about 90% of users can be clustered correctly, while maintaining the devices from only 0.7% of users may be clustered into the same cluster with others’ devices incorrectly.

The marine Internet of things (MIoT) is the application of the Internet of things technology in the marine field. Nowadays, with the arrival of the era of big data, the MIoT architecture has been transformed from cloud computing architecture to edge computing architecture. However, due to the lack of trust among edge computing participants, new solutions with higher security need to be proposed. In the current solutions, some use blockchain technology to solve data security problems while some use federated learning technology to solve privacy problems, but these methods neither combine with the special environment of the ocean nor consider the security of task publishers. In this article, we propose a secure sharing method of MIoT data under an edge computing framework based on federated learning and blockchain technology. Combining its special distributed architecture with the MIoT edge computing architecture, federated learning ensures the privacy of nodes. The blockchain serves as a decentralized way, which stores federated learning workers to achieve nontampering and security. We propose a concept of quality and reputation as the metrics of selection for federated learning workers. Meanwhile, we design a quality proof mechanism [proof of quality (PoQ)] and apply it to the blockchain, making the edge nodes recorded in the blockchain more high-quality. In addition, a marine environment model is built in this article, and the analysis based on this model makes the method proposed in this article more applicable to the marine environment. The numerical results obtained from the simulation experiments clearly show that the proposed scheme can significantly improve the learning accuracy under the premise of ensuring the safety and reliability of the marine environment.

Task scheduling on edge computing servers is an important issue that affects user experience. Existing scheduling methods require centralized control to achieve the best overall performance. However, it is impractical to force all users to act according to centralized control. We propose a distributed edge computing server task scheduling model based on game theory. Our method comprehensively considers the link quality from the mobile device to the server and the server's computing resource allocation when selecting edge computing servers, and achieves a balance between link quality and computing resources. Once the Nash equilibrium is reached, our model can provide different QoS for users of different priorities. Acceleration methods are proposed to achieve the Nash equilibrium faster. The simulation results show that the proposed model can provide differentiated services while optimizing the scheduling of computing resources, and ensure that the algorithm achieves an approximate Nash equilibrium in polynomial time.

WiFi-based human activity recognition has been widely used in many fields such as health diagnosis, intrusion detection and smart home. Most existing recognition methods can achieve a satisfying accuracy only in one domain, but low accuracy occurs when models are trained in source domain but are used in target domain. Meanwhile, considering finetuning network directly is impossible or easy to overfit with limited labeled target data, transfer learning based methods with domain adaptive layers are proposed to solve above problems but just aligning marginal distribution, which may lose massive fine-grained features. Based on this, we present an end-to-end deep subdomain adaptive network based activities recognition (DSANAR) using Channel State Information (CSI) that aligns marginal and matches conditional distribution simultaneously for more fine-grained features in each category of relevant subdomains based on a local maximum mean discrepancy (LMMD). Besides, by using a joint cross-entropy and an adaptive loss as training loss, DSANAR outperforms other state-of-art methods on an autonomous dataset with average 95.6% cross-domain accuracy.

Wi-Fi-based indoor localization gained a lot of attention over recent years due to low cost and open access properties. However, existing schemes might not be applicable in the real environment if their robustness is low. This paper presents CLRS, a novel distributed Indoor Positioning System (IPS) with high robustness which uses Wi-Fi signals to divide the space twice based on Angle of Arrival (AoA) and Effective Channel State Information (ECSI). The proposed scheme trade the redundancy of Access Point (AP) quantity to improve the tolerance of data measurement error. We performed simulations as well as real-world experiments, in which simulation results proved that the theoretical average error is the least when the routers are placed vertically in our localization method while the real-world experiments proved the high accuracy and robustness of CLRS.

Wireless networks that can exchange any type of data are vulnerable to multiple intrusions and increase potential security risks, so the design of an Intrusion Detection and Prevention System (IDPS) that analyzes the packet features and detects different intruders (i.e., the types of attack) is necessary. Whereas, the existence of redundant and irrelevant features hinders the potential of IDPS. In this paper, we propose TL-IDPS, a Two-Level classification IDPS of wireless network based on optimized features. In the phase of intrusion detection, one-hot method, normalization and correlation estimation are used to mitigate the redundant features. Then, the fuzzy membership function with cuttlefish algorithm maps and consolidates the extracted features and selects optimal features. Based on the optimal features, Di-distance k-nearest neighbor (K-NN) as the first level classify the intruder or non-intruder. Further the type of intruder is identified by deep Q-network. From the result of detected intruders, the further arrival of those intruders is prevented. Experimental results conducted from multiple evaluation metrics using the UNSW-NB15 dataset prove that our proposed TL-IDPS is more effective than existing IDPS methods.

Recent advances in Cross-Technology Communication (CTC) enable the coexistence and collaboration among heterogeneous wireless devices operating in the same ISM band (e.g., Wi-Fi, ZigBee, and Bluetooth in 2.4 GHz). However, state-of-the-art CTC schemes are vulnerable to spoofing attacks since there is no practice authentication mechanism yet. This paper proposes a scheme to enable the spoofing attack detection for CTC in heterogeneous wireless networks by using physical layer information. First, we propose a model to detect ZigBee packets and measure the corresponding Received Signal Strength (RSS) on Wi-Fi devices. Then, we design a collaborative mechanism between Wi-Fi and ZigBee devices to detect the spoofing attack. Finally, we implement and evaluate our methods through experiments on commercial off-the- shelf (COTS) Wi-Fi and ZigBee devices. Our results show that it is possible to measure the RSS of ZigBee packets on Wi-Fi device and detect spoofing attack with both a high detection rate and a low false positive rate in heterogeneous wireless networks.

The joint of WiFi‐based and vision‐based human activity recognition has attracted increasing attention in the human‐computer interaction, smart home, and security monitoring fields. We propose HuAc, the combination of WiFi‐based and Kinect‐based activity recognition system, to sense human activity in an indoor environment with occlusion, weak light, and different perspectives. We first construct a WiFi‐based activity recognition dataset named WiAR to provide a benchmark for WiFi‐based activity recognition. Then, we design a mechanism of subcarrier selection according to the sensitivity of subcarriers to human activities. Moreover, we optimize the spatial relationship of adjacent skeleton joints and draw out a corresponding relationship between CSI and skeleton‐based activity recognition. Finally, we explore the fusion information of CSI and crowdsourced skeleton joints to achieve the robustness of human activity recognition. We implemented HuAc using commercial WiFi devices and evaluated it in three kinds of scenarios. Our results show that HuAc achieves an average accuracy of greater than 93% using WiAR dataset.

WiFi-based Human activity recognition has attracted attention in the human-computer interaction, smart homes, and security monitoring fields. We first construct a WiFi-based activity dataset, namely WiAR, to provide a benchmark for existing works. Then, we leverage the moving variance of CSI to detect the start and end of activity. Moreover, we present K-means-based subcarrier selection mechanism according to subcarrier's sensitivity on human activity to enhance the robustness of human activity recognition. Finally, we leverage several classification algorithms to evaluate the performance of WiAR. Our results show that WiAR satisfies primary demand and achieves an average accuracy of greater than 93% using SVM, 80% using kNN, Random forest, and Decision tree.

A method for detecting human motion in complex scenarios based on Channel State Information (CSI) is presented. First, the sensitivity of CSI phase information to human motion is explored, especially to the strenuous motion. Through a large number of experiments, the influence of human motion on CSI phase is found out, and the characteristics of signal changes are extracted. The One-class Support Vector Machine (OSVM) in machine learning is used to detect the multi-target strenuous human motion. Line-Of-Sight (LOS) and Non-Line-Of-Sight (NLOS) conditions are studied in the case of obstacles appearing in the wireless link when human motion occurred. LOS and NLOS are identified by the skewness of the channel impulse response (CIR) distribution. After identifying the LOS condition and NLOS condition in the current environment, the human motion is analyzed and detected, which further improves the accuracy of human motion detection from 70% to 91%.

Recently, the low-power, low-cost and reliable ZigBee technology have received significant research attention with the increasing popularity of applications such as smart home system, patient monitor in hospitals. Coexisiting with the WiFi devices on the crowded unlicensed ISM band, such as hotspots and mobile phones, ZigBee will receive significant performance influence. The throughput and Packet Reception Rate (PRR) of ZigBee will decrease with the increasing number of WiFi devices. Because of the incompatible PHY/MAC layer, ZigBee devices will suffer near 50% packet loss when coexisting with WiFi devices. The existing mechanism such as CSMA/CA is surprisingly inadequate for solving this problem. As the WiFi traffic typically appears bursty, the channel will be free for more than 60% of the time. In this paper, we propose a new metric called channel idle indicator (CII) which can quantify the channel quality. Based on the CII and logistic regression, we build a channel idle state prediction model which can help ZigBee devices to use the white space of WiFi channel efficiently. Particularly, our approach is light-weight, which can be easily implemented on the ability-limited commercial off-the-shelf (COTS) ZigBee devices. Extensive experiments show that our scheme can achieve over 91% of the PRR, which is near 40% higher than the B-MAC protocol. When the WiFi throughput is 3Mbps, our scheme achieves near 1.5x throughput over B-MAC. Carrying on further, our scheme consumes less energy via degrading packet loss rate in the energy consumption part.

Currently, various Medium Access Control (MAC) protocols have been proposed for underwater Wireless sensor networks. Unlike terrestrial networks, underwater networks utilize acoustic waves, which have comparatively lower loss and longer range in underwater environments. However, the use of acoustic waves incurs long propagation delays that typically lead to low throughput especially in protocols that require receiver feedback such as multimedia stream delivery and the energy cost of transmission is much higher than receptions. Thus, collision and retransmission should be reduced in practice in order to reduce energy cost and improve throughput. Based on these motivations, we propose a novel MAC protocol called NCDC-MAC. NCDC-MAC leverages network coding and duty cycle, the combination of which is seldom explored, to solve these challenges. Heterogeneous wireless networks and node roles are considered while designing our algorithms. Meanwhile, fairness including schedule and service time assignment is supported in our approach. Extensive simulations show that our approach can achieve significantly better performance.

Many small businesses and public areas offer free Wi-Fi access, but may wish to restrict network access only to their customers or patrons inside the physical property. Unfortunately, due to the nature of wireless networks, this is difficult to accomplish. We develop and implement CLAC, a Crowdsourced Location aware Access Control scheme using physical layer information to address this challenge. It crowdsources both channel state information (CSI) and received signal strength (RSS) of already validated users to classify future users. We propose and use two CSI metrics in CLAC: CSI Cross-Antenna Stability Metric and CSI Cross-Frame Stability Metric, which summarize well the spatial and temporal CSI characteristics respectively. CLAC is evaluated in an office and a classroom. Evaluation results show that CLAC performs well in both environments, allowing most valid users inside the area to access the network, while the chance that invalid users outside the boundary may access the network is small.

Restricted by finite battery energy, traditional wireless sensor networks (WSNs) can only maintain for a limited period of time, resulting in serious performance bottleneck in long-term deployment of WSN. Fortunately, the advancement in the wireless energy transfer technology provides a potential to free WSNs from limited energy supply and remain perpetual operational. A mobile charger called wireless charging vehicle (WCV) is employed to periodically charge each sensor node and keep its energy level above the minimum threshold. Aiming at maximizing the ratio of the WCV's vocation time over the cycle time as well as guaranteeing the perpetual operation of networks, we proposes a feasible and optimal solution to this issue within the context of a real-time large-scale deployed WSN.

In recent years, Wi-Fi has seen extraordinary growth; however, due to the cost, performance and security issues, many Wi-Fi hotspot owners would like to restrict the network access only to individuals inside the physical property. Unfortunately, due to the nature of wireless, this is difficult to accomplish, especially with the off-the-shelf omni-antenna devices. In this work, we develop and implement CLaWa, a Crowdsourced Location Aware Wi-Fi Access Control scheme to address this challenge. Our system is based on observations of differing characteristics of physical layer information across physical boundaries such as walls and corners. CLaWa crowdsources both channel state information (CSI) and received signal strength (RSS) of already validated users to classify future users. We have also selected an appropriate machine learning algorithm for CLaWa. Evaluation results show that CLaWa can identify the boundary around a given area precisely, thus granting network access only to users inside the area while not validating users outside the boundary. Compared to indoor localization schemes, CLaWa is a lightweight solution which does not require expensive localization operations.

In recent years, the problem of spectrum mobility in Cognitive Radio (CR) Networks has been widely investigated. In order to fully utilize spectrum resources, many spectrum handoff techniques based on game theory have been proposed, but most studies only concern how to achieve better payoffs for users, without paying much attention to the Quality of Service (QoS). Thus, we propose a new channel switching model based on game theory, using a prioritized approach to meet the diverse needs of users, such as bandwidth, delay, and jitter. Once the Nash equilibrium is achieved, our model will provide different QoSes by setting different priorities to different users. We also propose two acceleration methods to reach the Nash equilibrium more quickly. Finally, we evaluate the performance of the proposed schemes using real channel availability measurements. Experiments results show that our model can provide differentiated services and our algorithm is guaranteed to reach an approximate Nash equilibrium within polynomial time.