The prevailing single-source paradigm in wireless sensing produces specialized models that are unscalable and generalize poorly to new tasks. Multi-source pre-training offers a path toward a generalist backbone but poses challenges including task heterogeneity, data redundancy, structural incompatibility, and the lack of a general-purpose pre-training objective. To address these issues, we propose WiSwiss, a comprehensive self-supervised multi-source pre-training framework that learns a general-purpose backbone for each modality. WiSwiss integrates semantic deduplication for dataset curation and a transformation-invariant pre-training objective. Experiments show that WiSwiss outperforms models trained from scratch, improving WiFi and mmWave performance by 4.5% and 10.3%, respectively, while reducing fine-tuning data requirements by 22.2% and 28.6%. We also present a qualitative study of scaling laws, showing that gains are task-dependent and that larger models require sufficiently large and diverse pre-training corpora to achieve substantial improvements.

Radio frequency (RF) fingerprinting exploits hardware imperfections for device identification, but distinguishing between same-model devices remains challenging due to their minimal hardware variations. Existing datasets for RF fingerprinting are constrained by small device scales and heterogeneous models, which hinder robust training and fair evaluation of machine learning methods. To address this gap, we introduce a large-scale dataset of same-model devices along with an open-source experimental framework. The dataset is built using 123 same-model commercial IEEE 802.11 g devices, which contain 35.42 million raw I/Q samples from the preambles and corresponding 1.85 million RF features. The accompanying framework further provides a fully reproducible pipeline from data collection to performance evaluation. Within this framework, a Random Forest–based algorithm is implemented as a baseline to achieve 88.6% identification accuracy on this dataset.

Industrial Internet of Things (IIoT) devices often rely on public communication channels for authentication, which makes them particularly susceptible to a variety of security threats. Physically Unclonable Function (PUF) provides a lightweight cryptographic method to address authentication challenges in resource-constrained devices. However, some PUF-based authentication frameworks store challenge-response pairs (CRPs) centrally on a server, posing risks of single point failure and performance bottlenecks in the authentication process. Although blockchain (BC) can mitigate this through decentralized storage, they still suffer from high computational overhead, large CRPs storage requirements, limited authentication efficiency, and ledger synchronization latency challenges. Therefore, we propose a novel direct mutual authentication framework for IIoT devices and peer nodes in consortium BCs, called PUF-D2PB. It enables peer nodes and IIoT devices to authenticate each other without involving BC-clients, and simultaneously supports CRP updates without requiring extra operations. In addition, we introduce a competition-based mechanism and a distributed encryption method to enhance CRP synchronization and prevent CRP leakage. Security analysis and a prototype implementation using a real PUF circuit demonstrate that our framework provides strong security guarantees and efficient authentication. Compared with existing BC-assisted PUF schemes, PUF-D2PB effectively reduces authentication latency and resource overhead while preventing CRP leakage.

Following the broad prospect of Radio Frequency Identification (RFID) technology is the security concern of unauthorized tag identification, which poses threats to the privacy of both objects and users. In this paper, we propose RF-Rock, the first RFID unauthorized identification attack that operates without tag activation, thereby evading almost all existing defenses. This attack exposes the vulnerabilities of current RFID networks in identification legitimacy and privacy. It is based on the intermodulation effect originating from intrinsic nonlinearity within tag circuits. To this end, we explore the distinctness and consistency of the intermodulation-based physical layer fingerprint of RFID tags with theoretical analysis and empirical validation, and optimize the attack accuracy and efficiency with delicate excitation plan. Real-world experiments show that RF-Rock achieves an attack success rate of 93.2% on average under various conditions. The entropy of our proposed fingerprint is 15.5 bits and implies sufficient capacity in practical attacks.

Radio frequency fingerprinting (RFF) is a technique that identifies wireless devices by exploiting unique hardware-induced imperfections measured in their transmitted RF signals. We propose a lightweight RFF method by leveraging temporal variations in RF signals. By combining multi-scale feature extraction through coarse and fine segmentation with DTW-based similarity features, our approach achieves an accuracy of 0.9882 and a Macro-F1 score of 0.988 in our experiment using the Wi-Fi RF data collected from 120 devices.

Deep learning-based object detection has seen a surge in applications for sensing systems on mobile devices. In this context, objects are identified and tracked across video frames, facilitating the calculation of associated events of interest. A significant research challenge refers to the acceleration of processing speed, which is constrained by deep learning-based object detection due to its intensive resource requirements. This paper focuses on a typical mobile sensing scenario, wherein sequences of frames containing objects of interest are sparsely dispersed throughout the video stream. Given that many of the frames lack objects, allocating substantial computational resources to detect them becomes inefficient. In light of this, we propose a stochastic scheduling algorithm, JumpQ. JumpQ performs per-frame detection when anticipating the presence of objects in the current frames. Consecutive negative detections prompt a transition to intermittent detection with a probability that undergoes further decay if the negative detection persists until reaching a predefined limit. Upon a positive detection, JumpQ swiftly reverts to per-frame detection and retraces a specific number of previously buffered frames to ensure the inclusion of potentially missed true frames. A comprehensive experimental study using the garbage bag counting technique was conducted to show the efficiency of JumpQ in accelerating the processing speed by nearly 1.92 times while maintaining a negligible impact on sensing accuracy.

The entropy source is the most critical component of a true random number generator (TRNG), which determines the quality of the random numbers. Current TRNGs mainly utilize a specific source of physical randomness as the entropy source, but it is difficult for this method to achieve a balance between low resource overhead and high throughput. This paper explores the self-feedback multiplexer (SFMUX) structure to obtain a novel dynamic hybrid entropy source for TRNGs. Unlike other MUX-based entropy source circuits, our SFMUX cross-connects the outputs of four independent high-frequency ring oscillators (ROs) as the input signals of four MUXs, and the output of each MUX is self-fed back to serve as a selection signal. Thus, the SFMUX can not only output jitter, but also update the selection signal rapidly and randomly, which increases the probability that the SFMUX outputs unstable signals. When using a D-flip-flop (DFF) to sample this signal, the DFF may become metastable. Modeling the entropy source shows that connecting 1-stage ROs and 2-stage ROs to each SFMUX can achieve higher minimum entropy than using ROs with other numbers of stages. The proposed TRNG design is implemented on Xilinx Virtex-6, Artix-7 and Kintex-7 FPGAs. The experimental results demonstrate that our TRNG achieves a maximum throughput of 550 Mbps while using only 6 slices, and it passes the NIST, AIS-31 and Dieharder tests without postprocessing.

One major research problem regarding mobile sensing systems is how to accelerate the processing speed which is bottle-necked by the deep-learning-based object detection. The focus of this paper is directed towards a typical mobile sensing scenario wherein sequences of frames containing interested objects are sparsely dispersed throughout the video stream. In light of this, we propose a stochastic scheduling algorithm named JumpQ. In the case of consecutive negative detections, JumpQ reduces the probability of detection, while in the case of positive detections, JumpQ promptly returns to frame-by-frame detection and retraces the buffered frames to detect the objects. Our experiment reports that the JumpQ algorithm accelerates processing speed by over 100%, all while incurring a negligible impact on sensing accuracy.

This paper demonstrates that thefractal dimension of frame preambleserves as a new radiometric feature that can be used together with other known radiometric features to enhance the identification accuracy in wireless device identification. We first propose a fractal dimension estimation scheme for direct-sequence spread spectrum (DSSS) frame preamble, then provide theoretical analysis to reveal how the fractal dimension is primarily determined by the device hardware imperfections, and thus prove that the fractal dimension serves as an intrinsic radiometric feature. We further show simulation results to verify our theoretical modeling of the fractal dimension and also numerically evaluate the effects of device hardware imperfections and wireless channels on the fractal dimension. Finally, by jointly applying the fractal dimension and the five features reported in the literature, we conduct extensive experiments to demonstrate that the fractal dimension can lead to a further improvement of the state-of-the-art result in the radiometric feature-based device identification.

Satellites serve as a key component for the upcoming space-air-ground integrated networks, while their signals are susceptible to spoofing attacks. Radio frequency fingerprinting (RFF) has been recognized as a highly promising authentication approach to counteracting spoofing attacks. Despite extensive RFF schemes proposed for terrestrial networks, RFF for satellites remains largely unexplored except for a recently reported scheme named PAST-AI, which exploits the spatial property of the IQ imbalance of downlink signals to authenticate Iridium satellites. Although PAST-AI demonstrates the potential of RFF for satellite authentication, its authentication accuracy and time are unsatisfactory. To address this issue, this paper proposes a novel 3D convolution-based RFF scheme for Iridium satellite authentication, which exploits not only the spatial property but also the temporal property of the IQ imbalance. The proposed RFF scheme transforms short-period sequences of successive IQ samples into 3D data samples and uses a 3D convolutional neural network (CNN) to train an RFF model. To evaluate the authentication accuracy, we collected over 198000000 IQ samples from all 66 Iridium satellites and generated 1000 3D data samples for each satellite. The results showed that the proposed RFF scheme achieves more accurate authentication than PAST-AI using fewer IQ samples (i.e., shorter time).

Crowdedness sensing of buses is playing an important role in the disease control of COVID-19 and bus resource scheduling. This research analyzes the relationship between carbon dioxide concentration, bus environment and the number of passengers by linear regression. Our prototype system collects the data of bus environment and carbon dioxide concentration to estimate the number of passengers in real time. By collecting the sensing data from a shuttle bus of university campus, we experimentally evaluate the feasibility and sensing performance of the crowdedness estimation model.

Traffic flows with different requirements of quality of service (QoS requirements) are aggregated into different QoS classes to provide differentiated services (Diffserv) and better quality of experience (QoE) for users. The existing aggregation approaches/QoS mapping methods are based on quantitative QoS requirements and static QoS classes. However, they are typically qualitative and time-varying at the edge of the beyond fifth generation (B5G) networks. Therefore, the artificial intelligence technology of preference logic is applied in this paper to achieve an intelligent method for edge computing, called the preference logic based aggregation model (PLM), which effectively groups flows with qualitative requirements into dynamic classes. First, PLM uses preferences to describe QoS requirements of flows, and thus can deal with both quantitative and qualitative cases. Next, the potential conflicts in these preferences are eliminated. According to the preferences, traffic flows are finally mapped into dynamic QoS classes by logic reasoning. The experimental results show that PLM presents better performance in terms of QoE satisfaction compared with the existing aggregation methods. Utilizing preference logic to group flows, PLM implements a novel way of edge intelligence to deal with dynamic classes and improves the Diffserv for massive B5G traffic with quantitative and qualitative requirements.

The information of daily garbage diposal can be used to develop many appealing applications in smart cities. This poster introduces DeepCounter, an automotive sensing system to providing a finegrained spatio-temporal distribution on the amount of disposed garbage bags. In the system, deep learning based image processing is used to automatically count the number of collected garbage bags from the video taken by a camera mounted on the rear of a garbage truck. A prototype system is implemented and experimental evaluation validates the feasibility of our proposal using realistic garbage collection videos in Fujisawa city Japan.

Collecting urban data in a citywide scale plays a fundamental role in the research, development and implementation of smart cities. This demo introduces Cruisers, an automotive sensing platform for smart cities, which is developed based on the following ideas. a) Garbage collecting trucks are used as host automobiles to accommodate sensors, b) 3G cellular communication network is used to wirelessly deliver sensed data directly to servers, and c) Proxy server(s) are adopted to convert the format of sensed data to required ones. This platform has been deployed to 24 garbage collecting trucks at Fujisawa city, i.e., nearly 1/4 of the total number of such trucks in the city. An iOS application is also developed to demonstrate the sensing process and the covered area.

Despite extensive efforts on exploring the asymptotic capacity bounds for mobile ad hoc networks (MANETs), the general exact capacity study of such networks remains a challenge. As one step to go further in this direction, this paper considers two classes of Aloha MANETs (A-MANETs) NAand NCthat adopt an aggressive traffic-independent Aloha and the conventional traffic-dependent Aloha, respectively. We first define a notation of successful transmission probability (STP) in NA, and apply queuing theory analysis to derive a general formula for the capacity evaluation of NA. We also prove that NCactually leads to the same throughput capacity as NA, indicating that the throughput capacity of NCcan be evaluated based on the STP of NA as well. With the help of the capacity formula and stochastic geometry analysis on STP, we then derive closed-form expressions for the throughput capacity of an infinite A-MANET under the nearest neighbor/receiver transmission policies. Our further analysis reveals that although it is highly cumbersome to determine the exact throughput capacity expression for a finite A-MANET, it is possible to have an efficient and closed-form approximation to its throughput capacity. Finally, we explore the capacity maximization and provide extensive simulation/numerical results.

A major obstacle stunting the application of mobile ad hoc networks (MANETs) is the lack of a general throughput capacity theory for such networks. Available works in this area mainly focused on exploring the order sense scaling laws of throughput capacity in MANETs with omnidirectional antennas or that of static ad hoc networks with directional antennas. Although the order sense results can help us to understand the general scaling behaviors, it tells us little about the exact throughput capacity. Another limitation of available works is that the impact of transmission power constraint on the throughput capacity is largely neglected. In most MANET applications, however, the mobile nodes are usually powered by batteries and have limited transmission power. In this paper, we study the exact throughput capacity of MANETs with directional antenna and transmission power constraint, where a generalized twohop relay algorithm with limited packet redundancy is adopted for packet routing. For given transmission power constraint, we first develop a model to map the omnidirectional transmission range to that of the directional one. We then explore the exact throughput capacity under directional transmission and group-based scheduling. Finally, numerical studies are provided to demonstrate the efficiency of these models and validate our theoretical results.