Timely and efficient dissemination of service information is critical in compute-first networking systems, where user requests arrive dynamically and computing resources are constrained. In such systems, the access point (AP) plays a key role in forwarding user requests to a server based on its latest received service information. This paper considers a single-source, single-destination system and introduces a PPO–based reinforcement learning framework for efficient information updating, guided by a newly designed reward metric called Age-and-Value-Aware (AVA). Unlike traditional freshness-based metrics, AVA explicitly incorporates variations in server-side service capacity and AP’s forwarding decisions, allowing more context-aware update evaluation. Under this reward structure, the PPO agent autonomously learns when to trigger service information updates, achieving a dynamic balance between communication cost and decision accuracy. Extensive simulations under diverse user request patterns and varying service capacities demonstrate that AVA reduces the update frequency by over 90% on average compared to baselines, with reductions reaching 98% in certain configurations. This reduction is achieved without compromising the quality of decision making.

In compute-first networking, computing router (or access points, APs) rely on edge service-status information to decide whether to offload tasks or process them locally. Under limited synchronization resources, maintaining fresher status at all APs does not necessarily improve decision quality, since many updates may not change the offloading outcome while still consuming scarce update budget. To address this mismatch, we propose the Error Decision Processing Time (EDPT) metric, which explicitly quantifies the processing-time penalty induced by erroneous offloading decisions and shifts status updating from freshness-oriented optimization to decision correctness. We then formulate the multi-AP status updating problem as a Markov Decision Process (MDP) with a per-slot update constraint and develop a Multi-Head Dueling Double DQN (MH-D3QN) framework. MH-D3QN employs AP-specific action-value estimation together with a benefit-ranking-guided action selection mechanism to mitigate the combinatorial action explosion and enable fast online decisions. Simulation results demonstrate that compared with the freshness-prioritized baseline, MH-D3QN saves more than 35.4% of the update budget and reduces the decision error rate by over 6.5%, effectively translating these gains into a 4.9% reduction in average task processing time.

LoRa holds significant promise for marine monitoring and communication due to its advantages of long range, low power consumption, and low cost. However, in marine environments, its communication performance is severely degraded by the strong absorption of electromagnetic waves by seawater. To enhance the reliability of communication under weak channels in the marine environment, this article proposes FCLoRa, a LoRa receiver enhancement system for low signal-to-noise ratio (SNR) marine environments. FCLoRa adopts a dual-domain cooperative strategy between the transmitter and receiver. On the receiver side, it employs a multilevel accumulation scheme to detect packets by aggregating the energy of windowed symbol “fragments” and reconstructs complete signals by fusing weak, fragmented signals from multiple gateways. On the transmitter side, it builds a two-dimensional polarization fingerprint library based on antenna attitude sensing and dynamically adjusts the transmission direction to match the polarization characteristics of the base station, thereby minimizing signal loss. Experimental results show that FCLoRa achieves a packet detection rate of nearly 40% at an extremely low SNR of –35 dB, and improves the average SNR by 1.92 dB compared to conventional LoRa reception, demonstrating its practical value in extreme marine scenarios.

As automotive voice assistants (AVAs) become increasingly cen- tral to modern vehicles, their vulnerability to attacks exploiting inaudible sounds should raise security concerns. However, such concerns are often deemed low priority, because it is widely be- lieved that an attacker to AVAs should be strategically positioned inside the concerned vehicle for two main reasons: i) inaudible signals can barely penetrate vehicle hulls and ii) a line-of-sight (LoS) path is needed between the attacker (sound source) and the AVA's microphone. In this paper, we disprove this common belief by proposing ShieldSpear to launch AVA attacks outside vehicle hulls. ShieldSpear exploits a tiny piezo-element placed on the exterior of the windshield to convert it into both a speaker and microphone. While this setting naturally brings the attacking sound source into a vehicle, strategically placing this compactly integrated element may further yield i) covertness (blended into stickers), ii) LoS path to AVA's microphones, and iii) real-time attacking capability dur- ing vehicle motion. To maintain sufficient volume while evading detection, we design novel hardware and signal carriers for deliver- ing attack (voice) commands. Moreover, ShieldSpear leverages the windshield-converted microphone to acquire drivers' voiceprint so as to accurately emulate it in the faked commands. Extensive experiments involving five mainstream vehicles have demonstrated the effectiveness of ShieldSpear by a 90.9% end-to-end success rate in injecting faked voice commands into AVAs.

The cross-boundary communication between unmanned aerial vehicles (UAVs) and autonomous underwater vehicles (AUVs) constitutes a pivotal component in achieving full coverage of the space-air-ground-sea integrated network of 6G. In such hybrid networks, it is imperative to intelligently plan the trajectory of UAVs by exploiting the deep reinforcement learning (DRL) technique and direct optical wireless communication (OWC) technology to ensure timely data collection. However, the traditional DRL technique suffers from issues such as sparse rewards and low sampling efficiency, which leads to a decrease in the freshness of data. To address these issues, in this paper, we investigate the trajectory planning problem for swarms of UAVs conducting data collection with age of information awareness. Leveraging the prior knowledge of large language models (LLMs) and multi-agent reinforcement learning technique, we propose a novel multi-UAV maritime data collection scheme. Firstly, we extract the prior knowledge strategies of LLMs in the form of expert trajectories through structured prompts. Then, initial strategy models are rapidly generated for UAVs through multi-agent behavior cloning method, which reduce ineffective exploration and accelerates learning. Finally, the MATD3 algorithm is used to fine-tune these strategy models, enhancing their policy learning capability in sparse reward environments. We also conduct simulations to validate the effectiveness of the proposed algorithms.

With the proliferation of smartphone application services, the pattern lock remains widely used for authentication. Notably, the risks associated with password entry in public spaces have attracted significant attention from researchers. Various attacks have been explored to steal passwords, but each comes with limitations, such as requiring good lighting conditions, close proximity, pre-deployed devices, or system intrusion. To address these challenges, we propose an attack method called BEyes, which utilizes beamforming feedback information (BFI) to eavesdrop on pattern passwords drawn on smartphone screens. Since BFI is transmitted in clear text and describes the downlink channel state information (CSI), any Wi-Fi 5-enabled device can capture it out of the victim’s view, reducing the likelihood of the attack being detected. To avoid missing critical pattern drawing information, we propose a traffic generation mechanism based on traffic competition, which ensures stable BFI. To mitigate the effects of frequency-selective fading and noise, we apply subcarrier alignment and principal component analysis (PCA) to improve efficiency. Additionally, we introduce a motion-based joint inference model, enabling BEyes to generalize its inference from a few known pattern passwords to unknown ones. Extensive experiments demonstrate that BEyes achieves an accuracy of 89.2% in inferring a 3-line pattern password within the Top-10 attempts.

In cooperative pursuit missions, multiple unmanned aerial vehicles (UAVs) can be leveraged to perform the pursuit task collaboratively. However, uneven energy distribution and low coordination efficiency among non-stationary agents lead to low pursuit success rate. To address these challenges, we propose a scheme which integrates energy load balancing with pursuit efficiency optimization leveraging multi-agent reinforcement learning approach. In this scheme, we first develop a CatBoost-based task allocation (CBTA) algorithm to balance multi-UAVs' residual energy and minimize overall pursuit time by designing an energy-efficiency scoring function which enables real-time online task allocation optimization. Then, to reduce policy fluctuations and improve optimal policy performance, we propose a FAA-MAAC algorithm performing multiple UAVs pursuit task. To deal with the low coordination efficiency problem, in this algorithm, we design a fluctuation-triggered asynchronous actor network update (FAAU) module and incorporate it into the original multi-actor attention critic (MAAC) framework to enhance policy stability and improve performance. Extensive simulation experiments are conducted to evaluate the scheme performance, and the simulation results show that the proposed scheme reduces pursuit time and improves success rates with three evasion strategies.

With the aging population, there is an increasing demand for rehabilitation services for people with chronic diseases. However, limitations such as medical resources, geographic barriers, and cost make home rehabilitation an option for more patients. Existing wearable devices and vision methods are effective but face problems with portability, cost, and privacy concerns. As for existing wireless sensing methods, they can only extract coarse features for activity recognition. Therefore, we present EchoHealth, which utilizes a smart speaker for rehabilitation exercise detection and assessment. We upgrade the smart speaker into an active sonar system without hardware modification to generate acoustic micro-distance images with motion information. Then, time-domain motion detection and distance-domain feature extraction are utilized to filter out the effects of non-motion time and distance to extract patient motion features for motion recognition. We further assess the patient's rehabilitation exercises from five aspects, based on which EchoHealth provides rehabilitation guidance. Extensive experiments with 15 participants performing 12 rehabilitation motions confirmed that EchoHealth can achieve 97.4% average accuracy in recognition of rehabilitation motion and provide accurate rehabilitation indicators in various environments.

Due to the complexity of ocean sensing tasks, buoy detection in traditional ocean observation methods has the disadvantages of high cost and insufficient real-time performance. Ocean mobile crowd sensing technology collects high-resolution data in real-time through shipborne sensors and has significant potential for future development in the ocean Internet of Things. In this article, we consider a cooperative vessel activity that performs sensing tasks, incentivizing multiple vessels through dynamic external task profits. To spontaneously incentivize the team to take part in tasks and maximize the total profit of the vessels, we propose a multiagent task allocation model (MATM-SN) based on deep reinforcement learning (DRL). This model uses a generative adversarial network to predict the trajectory of the vessels, and STGCN is used to extract synergistic features between vessels effectively. DRL training provides an optimal task allocation strategy for vessels. Finally, we compared four baselines and demonstrated that the MATM-SN model outperformed other baselines regarding task completion rate and total efficiency.

In recent years, with increasing user demands for convenience, privacy, and personalized experiences, gait recognition has been widely studied across various domains, such as indoor intrusion detection and smart homes. Although computer vision solutions are extensively researched for their visual intuitiveness, Wi-Fi sensing is emerging as a new research focus due to its ability to preserve privacy. However, previous studies have primarily relied on abstract features with limited interpretability or required multiple Wi-Fi links. To address these issues, we propose Wi-GR, which utilizes a Wi-Fi link to extract robust and highly interpretable gait features for user recognition. First, we construct a multi-path gait signal model to establish a clear relationship between Channel State Information (CSI) and gait motion. Then, we design a gait signal separation and enhancement method to mitigate the effects of external non-target reflections and internal multi-part reflections, which significantly impact the extraction and interpretability of gait features. Finally, fine-grained gait features that visualize gait patterns are generated using MUSIC-based and GAN-based multi-part velocity profile generation algorithms, tailored for single-person and multi-person scenarios, respectively. Numerous experiments have demonstrated that Wi-GR achieves single-person recognition accuracies of 95.3%, 94.0%, and 93.2% for 30 persons in the meeting room, corridor, and lobby, respectively, and an average accuracy of 88.3% for two-person recognition.

Password inference attacks by covert wireless side-channels jeopardize information safety, even for people with high security awareness and vigilance against snoopers. Yet, with limited spatial resolution, existing attacks cannot accurately infer password input on QWERTY keyboards in distance, creating psychological safety in using laptops publicly. To refute this false belief, we propose RefleXnoop, enabling an attacker to snoop a victim's typing details on a non-line-of-sight (NLoS) laptop. Apart from passively overhearing keystroke acoustic emanations, RefleXnoop actively probes with ultrasound, whose larger bandwidth and lower noise floor offers a finer resolution. To further maximize its performance, RefleXnoop exploits the laptop's screen reflection to enhance diversity in sound acquisition, and it innovates in neural models to effectively fuse the diversified sound acquisitions and to achieve robust feature-to-key translation. We implement RefleXnoop with commodity hardware and conduct extensive evaluation on it; the results demonstrate that RefleXnoop achieves 85% top-100 accuracy for inferring 8-character passwords on laptop QWERTY-keyboard and in multiple noisy environments.

In recent times, gait recognition, a type of biometric identification, has been widely used for area access control and smart homes. It improves convenience, privacy, and personalized experiences. Contemporary academic inquiry centers on privacy-preserving wireless sensing solutions as substitutes for computer vision. Yet, prevailing strategies heavily lean on abstract features, leading to inherent limitations in interpretability and stability. Fortunately, the widespread utilization of smart speakers has opened up opportunities for acoustic sensing, making it possible to extract more interpretable features. In this paper, we further push the limit of acoustic recognition with visual interpretability by sequentially visualizing fine-grained acoustic human gait features. The construction of initial gait profiles involves matrixing and compressing multipath gait echoes, resulting in imperceptible gait indications. Interpretability is then achieved through novel micro-range profiles, incorporating innovations such as clutter elimination using the Mobile Target Detector (MTD), compensation for farther echo strength, and subtraction of macro torso migration. These interpretable gait profiles offer practical benefits by enhancing data utilization, optimizing abnormal data handling, and improving model stability. Extensive evaluations with an open experimental scenario have been conducted to demonstrate accuracy reaching 97.5% in general, and robust performance against impacts from various practical factors.

Real-time indoor crowd counting plays an important role in many applications such as crowd control, resource allocation and advertisement. Current research predominantly relies on camera-based methods. However, computer vision-based solutions raise severe privacy and ethical concerns. In this paper, we propose a privacy-preserving counting solution called BeamCount based on Wi-Fi sensing. Instead of using conventional Wi-Fi Channel State Information (CSI) readings, we utilize Wi-Fi Beamforming Feedback Information (BFI) for crowd counting estimation. Compared to CSI which can only be extracted from few commodity Wi-Fi cards (e.g., Intel 5300), BFI readings can be obtained from a large range of commodity Wi-Fi devices. We establish a mapping relationship between BFI and headcount and extract headcounts from BFI inputs through a carefully designed adversarial network. Owing to the adversarial network's cross-domain capability, the proposed counting system can achieve high accuracy across different environments, demonstrating its generalization capability. To mitigate the effect of BFI compression on sensing performance, we adopt a novel time series prediction model. Extensive real-world experiments validate the effectiveness of BeamCount in various environments, achieving an average counting accuracy of 93.6%.

To balance venue safety with public photography rights, this article presents CamShield—a novel system for selective defense against unauthorized photography. Amid dense electromagnetic environments, CamShield reliably identifies cameras by analyzing their unintended electromagnetic emissions. By tracing frequency drift patterns and harmonic spectral movements unique to each device, CamShield can accurately detect cameras despite environmental noise or model similarities. An integrated antenna amplitude ratio module and Kalman filter further localize threats through resilient positioning. Directional ultrasonic beams then focus tuned acoustic interference toward devices, temporarily disrupting visualization in restricted locations while preserving ambient imaging freedoms. Comprehensive evaluations across three state-of-the-art object detectors quantify real-world reliability. With 30 intruding cameras, CamShield exhibited obstruction latencies below 346 ms. Furthermore, CamShield achieves three times the coverage using the same power as traditional Omnidirectional transmission. Together, the breakthroughs in pervasive camera sensing and context-aware actuation contribute toward advancing policy-centric access controls at the edge of cyber-physical convergence. CamShield sets an important precedent on enforcing venue custom protections in bounded secure zones without undermining positive public photography assumptions elsewhere.

Nowadays, sign language is becoming increasingly important in people’s daily life. Existing solutions are often based on wireless signals (e.g., acoustic, visible, and WiFi) or wearable sensors to recognize gestures, but they suffer from vulnerability to environmental influences, poor security, and high energy consumption, which prevent them from accurately capturing finger micromovements. In this article, we propose RF-Sign, which uses passive radio-frequency identification (RFID) tags to capture multiple finger micromovements simultaneously to enable sign language support. In particular, two main issues are studied. One is the problem of positional differences when users make the same gesture, and the other is the problem of segmenting consecutive gestures using only empirical thresholding methods and ignoring the existence of differences in thresholds for different gestures. For position differences, we propose position models to normalize the hand’s horizontal rotation angle and radial distance. For segmenting consecutive gestures, we use the received signal strength (RSS) trend of the reference tag to represent the finger micromovements state. The experimental results show that the average accuracy reaches 92.81% under different angles, distances, and other conditions.

Trip purpose is a meaningful aspect of travel behaviour for the understanding of urban mobility. However, it is non-trivial to automatically obtain trip purposes. On one hand, trip purposes are naturally diverse and complicated, but the available predictive data sources are limited in real-world scenarios. On the other hand, since trip purpose labeling is costly and the development levels of cities are unbalanced, it is infeasible to access large-scale labeled data in less developed cities to train advanced prediction models. To narrow the gaps, this article presents A new Graph Embedding Network and active Domain Adaptation based framework (AGENDA) that only requires open data sources and is capable of predicting in both label-rich cities and label-scarce cities. Specifically, in label-rich source cities, we first use the vehicle’s GPS trajectory and open POI check-ins to augment trip contexts. Then we establish a supervised graph embedding network with two attention mechanisms to extract the passenger’s latent activity semantics and a classifier to predict trip purpose. To enable the prediction in label-scarce target cities, we further devise an active domain adaptation framework, in which adversarial domain adaptation is used to transfer the source-learned knowledge, and active learning is used to integrate human intelligence in the model training. A group of experiments are conducted with real-world datasets in Beijing and Shanghai. Evaluation results demonstrate that the proposed framework significantly outperforms existing trip purpose prediction algorithms, and could make accurate trip purpose prediction in label-scarce cities with much fewer labeling efforts.

Acoustic target tracking has shown great advantages for device-free human-machine interaction over vision/RF-based mechanisms. However, existing approaches for portable devices solely track a single target, incapable of the ubiquitous and highly challenging multi-target situations such as double-hand multimedia controlling and multi-player gaming. In this paper, we proposeAMT$^+$, a pioneering smartphone MIMO system to achieve centimeter-level multi-target tracking. The challenge of multi-target occlusion is effectively addressed by employing multiple speaker-microphone pairs. However, the unique challenge raised by MIMO is the superposition of multi-source signals due to the cross-correlation among speakers. Initially, we tackle this challenge by designing a weak cross-correlation signal to reduce interference passively. InAMT$^+$, we’ve further integrated self-interference cancellation for active minimize interference. The most distinguishing advantage ofAMT$^+$lies in the elimination of the raised multipath effect, which is commonly ignored in previous work by hastily assuming targets as particles.AMT$^+$employs Doppler filtering over delay subtraction for echo suppression. Further, by non-particle target reflections modeling results, we introduce a distance-projection-based method for continuous target identification and tracking. Implemented on commercial smartphones,AMT$^+$achieves on average 0.54 cm, 1.37 cm, and 2.13 cm errors for single, double, and triple target tracking respectively, and on average 97.0% classification accuracy for 14 controlling gestures.

Recently, as fitness has become a popular part of people’s lives, the intention to record fitness processes and assess the standards of fitness movements has grown increasingly keen. However, the existing approaches have some limitations, for example, wearable devices can hinder users’ fitness activities; computer vision–based solutions pose the risk of privacy breach, and so on. Fortunately, we observed that smartspeaker, acoustic-based sensing is a promising method of activity monitoring. In this article, we propose Afitness, an acoustic-based sensing system that enables non-intrusive, passive, and high-precision fitness detection. Afitness has the following three innovations. (i) We utilize pulse compression to generate high-precision motion distance images on commercial devices that can be visually recognized. (ii) We propose a data augmentation algorithm, which also incorporates transfer learning to greatly reduce the pressure of data collection. (iii) We exploit incremental learning techniques that allow Afitness to improve the portability of our system and recognize new actions. Overall, Afitness achieves acoustic signal interpretability and environmental reliability detection.

The last decade has witnessed a tremendous growth of service computing, while efficient service recommendation methods are desired to recommend high-quality services to users. It is well known that collaborative filtering is one of the most popular methods for service recommendation based on QoS, and many existing proposals focus on improving recommendation accuracy, i.e., recommending high-quality redundant services. Nevertheless, users may have different requirements on QoS, and hence diversified recommendation has been attracting increasing attention in recent years to fulfill users’ diverse demands and to explore potential services. Unfortunately, the recommendation performances relies on a large volume of data (e.g., QoS data), whereas the data may be distributed across multiple platforms. Therefore, to enable data sharing across the different platforms for diversified service recommendation, we propose aPrivacy-preserving Diversified Service Recommendation(PDSR) method. Specifically, we innovate in leveraging the Locality-Sensitive Hashing (LSH) mechanism such that privacy-preserved data sharing across different platforms is enabled to construct a service similarity graph. Based on the similarity graph, we propose a novel accuracy-diversity metric and design a 2-approximation algorithm to select$K$services to recommend by maximizing the accuracy-diversity measure. Extensive experiments on real datasets are conducted to verify the efficacy of our PDSR method.

With the in-depth study of the internal process mechanism of the global ocean by oceanographers, traditional ocean observation methods have been unable to meet the new observation requirements. In order to achieve a low-cost ocean observation mechanism with high spatio-temporal resolution, this paper introduces mobile crowd sensing technology into the field of ocean observation. First, a Transformer-based vessel trajectory prediction algorithm is proposed, which can monitor the location and movement trajectory of vessel in real time. Second, the participant selection algorithm in mobile crowd sensing is studied, and based on the trajectory prediction algorithm, a dynamic participant selection algorithm for ocean mobile crowd sensing is proposed by combining it with the discrete particle swarm optimization (DPSO) algorithm. Third, a coverage estimation algorithm is designed to estimate the coverage of the selection scheme. Finally, the spatio-temporal resolution of the vessel's driving trajectory is analyzed through experiments, which verifies the effectiveness of the algorithm and comprehensively confirms the feasibility of mobile crowd sensing in the field of ocean observation.

Digital Twins (DTs) can generate digital replicas for mobile networks (MNs) that accurately reflect the state of MN. Machine learning (ML) models trained in DT for MN (DTMN) virtual environments can be more robustly implemented in MN. This can avoid the training difficulties and runtime errors caused by MN instability and multiple failures. However, when using data from various devices in the MN system, DTs must prioritize data privacy. Federated learning (FL) enables the construction of models without data leaving devices to protect DTMN data privacy. Nevertheless, FL’s privacy protection needs further improvement for it only guarantees device-level data ownership but ignores that models may retain private information from data. Therefore, this paper focuses on data forgetting in privacy protection, and proposes a novel FL-based unlearning framework (FedME2), which contains MEval and MErase modules. Guided by memory evaluation information from MEval and employing MErase’s multi-loss training approach, FedME2 gets accurate data forgetting in DTMN. In four DTMN virtual environments, FedME2 achieves an average data forgetting rate of approximately 75% for global models under FL and kept the influence on global models’ accuracy below 4%. FedME2 has better data forgetting and improves DTMN data privacy protection while guaranteeing model accuracy.

Acoustic target tracking has shown significant potential for contactless human-computer interaction. However, most existing acoustic 2-D tracking approaches for portable devices require at least one speaker and two microphones, incapable for universal devices. In this paper, we propose NearTracker, a contactless acoustic tracking system, achieves 2-D target tracking with only one speaker and one microphone (i.e., Single Input Single Output, SISO). With the help of a nearby reflector, the additional valuable echoes from target are combined for positioning. Actually, the dynamic interferences from non-target echoes pose huge challenges for target echo extraction. NearTracker extracts and enhances these faint target echoes with novel signal processing methods and estimates the target’s location accurately via a designed particle filter algorithm. Extensive experiments show that our system achieves on average 1.36 cm error for 2-D target tracking, which can satisfy most devices and application scenarios.

A practical challenge common to state-of-the-art acoustic gesture recognition techniques is to adaptively respond to intended gestures rather than unintended motions during the real-time tracking on human motion flow. Besides, other disadvantages of under-expanded sensing space and vulnerability against mobile interference jointly impair the pervasiveness of acoustic sensing. Instead of struggling along the bottlenecked routine, we innovatively open up an independent sensing dimension of acoustic 2-D hand-shape imaging. We first deductively demonstrate the feasibility of acoustic imaging through multiple viewpoints dynamically generated by hand movement. Amaging, hand-shape imaging triggered gesture recognition, is then proposed to offer adaptive gesture responses. Digital Dechirp is novelly performed to largely reduce computational cost in demodulation and pulse compression. Mobile interference is filtered by Moving Target Indication. Multi-frame macro-scale imaging with Joint Time-Frequency Analysis is performed to eliminate image blur while maintaining adequate resolution. Amaging features revolutionary multiplicative expansion on sensing capability and dual dimensional parallelism for both hand-shape and gesture-trajectory recognition. Extensive experiments and simulations demonstrate Amaging’s distinguishing hand-shape imaging performance, independent from diverse hand movement and immune against mobile interference. 96% hand-shape recognition rate is achieved with ResNet18 and 60× augmentation rate.

Approximate Nearest Neighbor(ANN) search is the core problem in many large-scale machine learning and computer vision applications such as multimodal retrieval. Hashing is becoming increasingly popular, since it can provide efficient similarity search and compact data representations suitable for handling such large-scale ANN search problems. Most hashing algorithms concentrate on learning more effective projection functions. However, the accuracy loss in the quantization step has been ignored and barely studied. In this paper, we analyse the importance of various projected dimensions, distribute them into several groups and quantize them with two types of values which can both better preserve the neighborhood structure among data. One is Variable Integer-based Quantization (VIQ) that quantizes each projected dimension with integer values. The other is Variable Codebook-based Quantization (VCQ) that quantizes each projected dimension with corresponding codebook values. We conduct experiments on five common public data sets containing up to one million vectors. The results show that the proposed VCQ and VIQ algorithms can both achieve much higher accuracy than state-of-the-art quantization methods. Furthermore, although VCQ performs better than VIQ, ANN search with VIQ provides much higher search efficiency.

The reduction in the number of queries to the object model is a hot topic in the current research of black-box adversarial attack methods. To solve this problem, in this article, we propose generating adversarial examples with shadow model (GASM) that shifts the number of queries to the object model to the shadow model. The method first determines the shadow model based on the robustness and transferability of classifiers and fine-tunes the decision boundary of the shadow model by constructing adversarial datasets. Second, accesses the shadow model and constructs adversarial examples by maximizing the output probability of the targeted class (any class other than the current one) to modify the image gradient information. Finally, the results show that GASM has the strongest transferability and outperforms white-box attacks when AlexNet (MNIST), VGG-19 (CIFAR10), and MobileNet v2 (Tiny ImageNet) are selected as shadow models.

Hashing has been widely applied to the large-scale approximate nearest neighbor search problem owing to its high efficiency and low storage requirement. Most investigations concentrate on learning hashing methods in a centralized setting. However, in existing big data systems, data is often stored across different nodes. In some situations, data is even collected in a distributed manner. A straightforward way to solve this problem is to aggregate all the data into the fusion center to obtain the search result (aggregating method). However, this strategy is not feasible because of the prohibitive communication cost. Although a few distributed hashing methods have been proposed to reduce this cost, they only focus on designing a distributed algorithm for a specific global optimization objective without considering scalability. Moreover, existing distributed hashing methods aim at finding a distributed solution to hashing, meanwhile avoiding accuracy loss, rather than improving accuracy. To address these challenges, we propose a Scalable Distributed Hashing (SDisH) model in which most existing hashing methods can be extended to process distributed data with no changes. Furthermore, to improve accuracy, we utilize the search radius as a global variable across different nodes to achieve a global optimum search result for every iteration. In addition, a voting algorithm is presented based on the results produced by multiple iterations to further reduce search errors. Theoretical analyses of communication, computation, and accuracy demonstrate the superiority of the proposed model. Numerical simulations on three large-scale and two relatively small benchmark datasets also show that the SDisH model achieves up to 44.75% and 10.23% accuracy gains compared to the aggregating method and state-of-the-art distributed hashing methods, respectively.

Acoustic target tracking has shown great advantages for device-free human-machine interaction over vision/RF based mechanisms. However, existing approaches for portable devices solely track single target, incapable for the ubiquitous and highly challenging multi-target situation such as double-hand multimedia controlling and multi-player gaming. In this paper, we propose AMT, a pioneering smartphone MIMO system to achieve centimeter-level multi-target tracking. Targets' absolute distance are simultaneously ranged by performing multi-lateration locating with multiple speaker-microphone pairs. The unique challenge raised by MIMO is the superposition of multisource signals due to the cross-correlation among speakers. We tackle this challenge by applying Zadoff-Chu(ZC) sequences with strong auto-correlation and weak cross-correlation. The most distinguishing advantage of AMT lies in the elimination of target raised multipath effect, which is commonly ignored in previous work by hastily assuming targets as particles. Concerning the multipath echoes reflected by each non-particle target, we define the novel concept of primary echo to best represent target movement. AMT then improves tracking accuracy by detecting primary echo and filtering out minor echoes. Implemented on commercial smartphones, AMT achieves on average 1.13 cm and 2.46 cm error for single and double target tracking respectively and on average 97% accuracy for 6 controlling gestures recognition.

Short-term prediction of fishing effort distributions will guide fishery management in a dynamic way. However, it meets two unique challenges: the randomness of fishers’ behaviors and the diversity of marine meteorology such as sea storms in the short period. This study proposes short-term prediction system of fishing effort distribution by mining a new kind of knowledge: fishing chronology among trawlers. We first define, quantify, and dig out chronological fishing relations among trawlers based on the VMS dataset. Then the system extracts the optimal early bird set from chronological fishing relations, whose current fishing behaviors can serve as indicators of future fishing effort distributions. Based on the knowledge of fishing chronology, we further design a Convolution Neural Network (CNN) to predict the short-term fishing effort distribution, only taking the current fishing behaviors of early birds as input. We evaluate the system performance on the VMS dataset of 1589 trawlers in the East China Sea from October 2015 to April 2017. The system uses the VMS traces in the first half period to calculate fishing chronology among trawlers, to extract early birds, and to train the CNN model. The traces in the last half period is used to evaluate the prediction accuracy. The results confirm a low prediction error ratio of 6.95% across all the weeks only by tracking 19 early birds. More importantly, our prediction system keeps its accuracy during the week of a sea storm in Feb. 8th to 10th, 2017. The application of our system for fishery management is encouraging: tracking only 1% trawlers suffices to predict short-term fishing effort distributions in the near future.

Hashing methods have been widely used in Approximate Nearest Neighbor (ANN) search for big data due to low storage requirements and high search efficiency. These methods usually map the ANN search for big data into the k -Nearest Neighbor ( k NN) search problem in Hamming space. However, Hamming distance calculation ignores the bit-level distinction, leading to confusing ranking. In order to further increase search accuracy, various bit-level weights have been proposed to rank hash codes in weighted Hamming space. Nevertheless, existing ranking methods in weighted Hamming space are almost based on exhaustive linear scan, which is time consuming and not suitable for large datasets. Although Multi-Index hashing that is a sub-linear search method has been proposed, it relies on Hamming distance rather than weighted Hamming distance. To address this issue, we propose an exact k NN search approach with Multiple Tables in Weighted Hamming space named WHMT, in which the distribution of bit-level weights is incorporated into the multi-index building. By WHMT, we can get the optimal candidate set for exact k NN search in weighted Hamming space without exhaustive linear scan. Experimental results show that WHMT can achieve dramatic speedup up to 69.8 times over linear scan baseline without losing accuracy in weighted Hamming space.

An efficient and low‐cost communication system has great significance in maritime communication, but it faces enormous challenges because of high communication costs, incomplete communication infrastructure, and inefficient routing algorithms. Delay Tolerant Vessel Networks (DTVNs), which can create low‐cost communication opportunities among vessels, have recently attracted considerable attention in the academic community. Most existing maritime ad hoc routing algorithms focus on predicting vessels’ future contacts by mining coarse‐grained social relations or spatial distribution, which has led to poor performance. In this paper, we analyze 3‐year trajectory data of 5123 fishery vessels in the China East Sea. Using entropy theory, we observe that the trajectory of the vessel has strongly spatial‐temporal distribution regularity, especially when previous states were given. To predict accurate future trajectories, we develop a long‐term accurate trajectory prediction model by improving the Bidirectional Long‐Short Term Memory (Bi‐LSTM) model. Based on predicted trajectories and the confident degree of each prediction step, we propose a series of routing algorithms called TPR‐DTVN to achieve efficient communication performance. Finally, we carry out simulation experiments with extensive real data. Compared with existing algorithms, the simulation results show that TPR‐DTVN can achieve a higher delivery ratio with lower cost and transmission delay.

Internet of Things realizes the leap from traditional industry to intelligent industry. However, it makes edge devices more vulnerable to attackers during processing perceptual data in real time. To solve the above problem, we use the zero‐sum game to build the interactions between attackers and edge devices and propose an antiattack scheme based on deep reinforcement learning. Firstly, we make the k NN‐DTW algorithm to find a sample that is similar to the current sample and use the weighted moving mean method to calculate the mean and the variance of the samples. Secondly, to solve the overestimation problem, we develop an optimal strategy algorithm to find the optimal strategy of the edge devices. Experimental results prove that the new scheme improves the payoff of attacked edge devices and decreases the payoff of attackers, thus forcing the attackers to give up the attack.

Freehand exercises help improve physical fitness without any requirements on devices, or places (e.g., gyms). Existing fitness assistant systems require wearing smart devices or exercising at specific positions, which compromises the ubiquitous availability of freehand exercises. This work proposes MobiFit, a contactless freehand exercise assistant using just one cellular signal receiver. MobiFit monitors the ubiquitous cellular signals sent by the base station and provides accurate repetition counting, exercise type recognition, and workout quality assessment without any attachments to the human body. To design MobiFit, we first analyze the characteristics of the received cellular signal sequence during freehand exercises through experimental studies. Based on the observation, we construct the analytic model of the received signals. Guided by the analytic model, MobiFit segments out every repetition and rest interval from one exercise session through spectrogram analysis, and extracts low-frequency features from each repetition for type recognition. We have implemented the prototype of MobiFit and collected 22,960 exercise repetitions performed by ten volunteers over six months. The results confirm that MobiFit achieves high counting accuracy of 98.6%, high recognition accuracy of 94.1%, and low repetition duration estimation error within 0.3s. Besides, the experiments show that MobiFit works both indoor and outdoor, and supports multiple users exercising together.

The construction of smart cities is the concentrated embodiment of the application of Internet of Things (IoT), which provides a variety of solutions to various problems faced by urban development and urban management. However, the openness, dynamics, and heterogeneity of city IoT increase the risk of attacks, especially the social attributes contained in such systems accelerate virus propagation. To reduce the risk and harm of virus propagation, it is of great significance to analyze the mode and the characteristics of virus propagation. This article proposes a dynamic virus propagation model [i.e., Ignorant-DKs- HN-Exposed-Pvirus-Spread-Recover (IDEPSR)], which focuses on two social attributes (i.e., intelligent device's propagation capability and identification ability). First, this article presents a new algorithm (i.e., DKs-HN) to measure the first attribute based on the evidence theory and an improved K-shell method. The DKs- HN merges the direct influence of a particular node and the indirect influence of 1-hop neighbors by utilizing the combination rules of the evidence theory. Second, this article proposes a Pvirus method to measure the second attribute based on the social hierarchy theory and the nonnegative matrix factor method. Every intelligent device can identify its trust relationship with other devices by using the Pvirus, and then the probability of virus activation lurking in devices can be predicted. Finally, a real data set is used to simulate a social-aware city IoT environment. Convincing experimental results show that the IDEPSR is more reasonable in design and good in performance. The IDEPSR performs better in controlling the virus propagation than the other five models.

Under the framework of multi-label classification, the excessive training time restricts the availability of non-linear kernel SVM (Support Vector Machine) classification algorithm on large-scale data sets. To solve this problem, this paper provides a fast multi-label SVM classification algorithm b ased on approximate extreme points (AEMLSVM). Firstly, it utilizes the approximate extreme point technique to obtain representative sets from the training data set. These representative sets not only retain almost all information of the training data set, but also its size is much smaller than that of training data set. After that, SVM is trained on the representative sets. Furthermore, the improved AEMLSVM algorithm (AEMLSVM-DEC) adopts DEC (Different Error Costs) technique to solve the label data imbalanced problem. We have conducted extensive experiments on four large-scale benchmark data sets. The results show that the proposed algorithms can effectively reduce training time, and their classification performance is similar to that of the traditional multi-label SVM algorithm. They outperform other scalable multi-label SVM algorithms in training time and classification performance. By adopting DEC method to solve the label data imbalanced problem, the AEMLSVM-DEC algorithm has a better classification performance than AEMLSVM algorithm.