Digital Twin (DT) technology serves as a critical enabler in Cyber-Physical-Social Systems (CPSS), especially within Industry 5.0’s human-centric manufacturing paradigm. However, the computational intensity of processing real-time data in DT systems often leads to resource saturation and performance degradation at computation nodes. Dynamic service migration of digital twins by offloading computation-intensive tasks to resource-rich nodes to offer a promising solution, yet introduces challenges in preserving service quality during migration. Key issues include high delay, data inconsistency, service interruptions, and limited bandwidth compromising system stability. To address these challenges, this paper proposes a dynamic migration scheduling strategy for digital twins based on a Lyapunov optimization framework. Our approach integrates Stochastic Network Calculus (SNC) for Quality of Service (QoS) quantification and a persistent queue mechanism for reliability assurance. Theoretical analysis and extensive simulations demonstrate that the proposed algorithm achieves near-optimal performance with provable bounds, effectively minimizing migration-induced delay while maintaining service reliability. The results confirm that our framework consistently outperforms existing solutions in managing service migration within industrial internet systems.

With the widespread application of machine learning-based graph classification models in fields such as biology and economics, there has been a growing number of attacks aimed at perturbing classification results. Although current defense methods, such as randomized smoothing, have achieved some success, their practical applicability remains limited due to the need to modify classification models to ensure accuracy.In this paper, we propose a novel defense method—GRACED, which provides theoretical guarantees for the accuracy and robustness of graph classification without requiring knowledge of the attacker’s capabilities or the classification model. The key idea behind our method is to leverage the denoising ability of feature diffusion models for adversarial data purification. We then demonstrate that this randomized purification approach can ensure certified robustness under specific attack budgets. Extensive experiments confirm our theoretical findings and show that graph classifiers using GRACED significantly outperform state-of-the-art classifiers. For instance, the accuracy on MUTAG improved by 11%, and the best results on IMDB showed a 14% increase.

Recently image steganography methods based on invertible neural networks (INNs) demonstrated the capability to automatically embed and extract secret messages while maintaining high visual quality in stego images. However, there remain concerns about security and invertibility of such methods. In this paper, for the first time, we introduce adversarial hiding into INN-based image steganography method to simultaneously perform steganographic embedding and adversarial perturbation generation, resulting in improved security. Our method enhances the invertibility of the INN structure: It utilizes the lost information of the INN to generate perturbations, which are then combined with the gradient of the cover image to produce an adversarial stego image. Also, a learnable noise layer is proposed to mitigate information loss caused by rounding and truncation during image storage. Therefore, the proposed method significantly improves security while enhancing extraction performance of INN-based steganography approach, as supported by our experimental results. For example, the steganalysis detection accuracy of SRNet decreases from 96.86% to 51.77% at a payload of 0.2 bits per pixel (bpp).

Environmental sensing is an important research topic in the integrated sensing and communication (ISAC) system. Current works often focus on static environments, such as buildings and terrains. However, dynamic factors like rainfall can cause serious interference to wireless signals. In this paper, we propose a system called RainfalLTE that utilizes the downlink signal of LTE base stations for device-independent rain sensing. In particular, it is fully compatible with current communication modes and does not require any additional hardware. We evaluate it with LTE data and rainfall information provided by a weather radar in Badaling Town, Beijing The results show that for 10 classes of rainfall, RainfalLTE achieves over 97 % identification accuracy. Our case study shows that the assistance of rainfall information can bring more than 40 % energy saving, which provides new opportunities for the design and optimization of ISAC systems.

In recent years, image steganography methods based on invertible neural networks (INNs) have received significant attention due to their invertible structure, which offers advantages in embedding and extracting secret messages. However, current INN-based image steganography methods face challenges, particularly their limited tolerance against noise interference (e.g., added Gaussian noise, adversarial perturbations, and JPEG compression) and vulnerability to detection by advanced deep steganalyzers. To address these concerns, we present a novel steganography framework that combines Median Smoothing Training (MST) with dynamic Projected Gradient Descent (d-PGD). Specifically, our method begins with employing an MST strategy during the training phase to improve the INN’s tolerance to noise, ensuring that accurate message extraction even under noise interference. Subsequently, to improve the security of INN-based steganography, we propose a d-PGD algorithm that can generate minimal adversarial perturbations capable of deceiving deep steganalyzers, thereby improving security without compromising extraction accuracy. Experimental results demonstrate that our method achieves state-of-the-art secret message extraction accuracy while significantly improving resistance against deep steganalyzers.

A great challenge to steganography has arisen with the wide application of steganalysis methods based on convolutional neural networks (CNNs). To this end, embedding cost learning frameworks based on generative adversarial networks (GANs) has been proposed and achieved success for spatial image steganography. However, the application of GAN to JPEG steganography is still in the prototype stage; its antidetectability and training efficiency should be improved. In conventional steganography, research has shown that the side information calculated from the precover can be used to enhance security. However, it is hard to calculate the side information without the spatial domain image. In this work, an embedding cost learning framework for JPEG image steganography via a GAN (JS–GAN) has been proposed, the learned embedding cost can be further adjusted asymmetrically according to the estimated side information (ESI). Experimental results have demonstrated that the proposed method can automatically learn a content‐adaptive embedding cost function, and using the ESI properly can effectively improve the security performance. For example, under the attack of a classic steganalyzer GFR with a quality factor of 75 and 0.4 bpnzAC, the proposed JS–GAN can increase the detection error by 2.58% over J‐UNIWARD, and the ESI–aided version JS–GAN (ESI) can further increase the security performance by 11.25% over JS–GAN.

The effectiveness of WiFi-based localization systems heavily relies on the spatial accuracy of WiFi AP. In real-world scenarios, factors such as AP rotation and irregular antenna tilt contribute significantly to inaccuracies, surpassing the impact of imprecise AP location and antenna separation. In this paper, we proposeAnteumbler, a non-invasive, accurate, and efficient system for measuring the orientation of each antenna in physical space. By leveraging the fact that maximum received power occurs when a Tx-Rx antenna pair is perfectly aligned, we build a spatial angle model capable of determining antennas’ orientations without prior knowledge. However, achieving comprehensive coverage across the spatial angle necessitates extensive sampling points. To enhance efficiency, we exploit the orthogonality of antenna directivity and polarization, and adopt an iterative algorithm, thereby reducing the number of sampling points by several orders of magnitude. Additionally, to attain the required antenna orientation accuracy, we mitigate the influence of propagation distance using a dual plane intersection model while filtering out ambient noise. Our real-world experiments, covering six antenna types, two antenna layouts, two antenna separations ($\lambda /2$and$\lambda$), and three AP heights, demonstrate thatAnteumblerachieves median errors below$\text{6}^\circ$for both elevation and azimuth angles, and exhibits robustness in NLoS and dynamic environments. Moreover, when integrated into the reverse localization system,Anteumblerdeployed over LocAP reduces antenna separation error by$10 \,\mathrm{mm}$, while for user localization system, its integration over SpotFi reduces user localization error by more than$1 \,\mathrm{m}$.

WiFi-based gait recognition technologies have seen significant advancements in recent years. However, most existing approaches rely on a critical assumption: users must walk continuously and maintain a consistent body posture. This poses a substantial challenge when users engage in non-periodic or discontinuous behaviors (e.g., stopping, starting, or turning mid-walk), which can disrupt the extraction of gait-related features and degrade recognition performance. To address this issue, we proposefreeGait, a novel approach designed to mitigate the impact of non-gait behaviors in WiFi-based gait recognition systems. Our solution models this problem as domain adaptation, where we learn domain-independent representations to isolate gait features from behavior-dependent noise. We treat human behaviors with labeled user data as source domains and behaviors without user labels as target domains. However, applying domain adaptation directly is challenging due to the ambiguous classification boundaries in the target domains for WiFi signals. To overcome this, we align the posterior distributions between the source and target domains and constrain the conditional distribution within the target domains to enhance gait classification accuracy. Additionally, we implement a data augmentation module to generate data resembling the labeled data, while supervised learning ensures distinctiveness between users. Our experiments, conducted with 20 participants across 3 different scenarios, demonstrate thatfreeGaitcan accurately predict data across 15 domains by labeling only a small subset from 6 source domains, achieving up to a 45% improvement in user classification accuracy compared to existing methods.

Accurately estimating the velocity (including speed and direction) of moving targets has recently attracted widespread attention in augmented reality, security monitoring and sports health. In particular, the Doppler Frequency Shift (DFS)-based velocity estimation schemes using WiFi devices have shown great potential and have been widely studied. However, previous Fast Fourier Transform (FFT)-based and path-parameter-based arts have inherent limitations in DFS estimation and, worse still, ignore the nonlinear measurement errors caused by the relative orientation between the moving target and the WiFi transceiver. The above limitations make it difficult to meet the requirements for fine-grained velocity estimation in practical applications. To cope with these limitations, in this article, we propose a learning-based velocity estimation framework, named freeDoppler , to achieve fine-grained, multi-target and orientation-independent velocity estimation. Specifically, we construct a WiFi-based Velocity Estimation Network (VEN), which leverages continuous complex-valued Channel State Information (CSI) sequences as input, to fully learn the inherent information of the Doppler effect and accurately predict velocity series. In addition, we adopt the electric field scattering model of Maxwell’s equations to construct a physics-informed CSI Generation Model (CGM), thereby generating large-scale and high-quality simulated CSI samples to improve the generalization of the VEN model. Throughout extensive real-world experiments, freeDoppler can achieve median errors of 7.98 cm/s for speed estimation, 28° for direction estimation and 35 cm for human tracking in one or two moving targets, significantly outperforming the state-of-the-art methods.

The performance of WiFi-based localization systems is affected by the spatial accuracy of WiFi AP. Compared with the imprecision of AP location and antenna separation, the imprecision of AP’s or antenna’s orientation is more important in real scenarios, including AP rotation and antenna irregular tilt. In this paper, we propose Anteumbler that non-invasively, accurately and efficiently measures the orientation of each antenna in physical space. Based on the fact that the received power is maximized when a Tx-Rx antenna pair is perfectly aligned, we construct a spatial angle model that can obtain the antennas’ orientations without prior knowledge. However, the sampling points of traversing the spatial angle need to cover the entire space. We use the orthogonality of antenna directivity and polarization and adopt an iterative algorithm to reduce the sampling points by hundreds of times, which greatly improves the efficiency. To achieve the required antenna orientation accuracy, we eliminate the influence of propagation distance using a dual plane intersection model and filter out ambient noise. Our real-world experiments with six antenna types, two antenna layouts and two antenna separations show that Anteumbler achieves median errors below 6 ° for both elevation and azimuth angles, and is robust to NLoS and dynamic environments. Last but not least, for the reverse localization system, we deploy Anteumbler over LocAP and reduce the antenna separation error by 10 mm, while for the user localization system, we deploy Anteumbler over SpotFi and reduce the user localization error by more than 1 m.

Recently, WiFi-based gait recognition technologies have been widely studied. However, most of them work on a strong assumption that users need to walk continuously and periodically under a constant body posture. Thus, a significant challenge arises when users engage in non-periodic or discontinuous behaviors (e.g., stopping and going, turning around during walking). This is because variations of non-gait behaviors interfere with the extraction of gait-related features, resulting in recognition performance degradation. To solve this problem, we propose freeGait, which aims to mitigate the user's non-gait behaviors of WiFi-based gait recognition system. Specifically, we model this problem as domain adaptation, by learning domain-independent representations to extract behavior-independent gait features. We consider human behaviors with labels of users as source domains, and human behaviors without labels of users as target domains. However, directly applying domain adaptation to our specific problem is challenging, because the classification boundaries of the unknown target domains are unclear for WiFi signals. We align the posterior distributions of the source and target domains, and constrain the conditional distribution of the target domains to optimize the gait classification accuracy. To obtain enough source domains data, we build a data augmentation module to generate data similar to the labeled data, and use supervised learning to make the data different between users. We conduct experiments with 20 people and 3 different scenarios, and the results show that accurate predictions of a total of 15 domains data can be achieved by only collecting and labeling a small amount of data from 6 source domains, and user classification accuracy can be improved by up to 45% compared to other existing techniques.

Accurate and quick relative-distance measurement is crucial for supporting various intelligent transparent services, such as multi-device collaboration, screen rotation, and multi-device mirroring. Unfortunately, current methods often rely on single-modality sensing, resulting in various limitations: BLE-based and WiFi-based methods suffer from coarse-grained estimation, and ultrasound-based approaches suffer from limited sensing range. In this work, we aim at designing a distance-measurement system that enjoys long range, high accuracy, and small delay. Our designed system, named BAIR, relies on low-energy Bluetooth (BLE), acoustic sensors, and inertial measurement units (IMU) equipped on commercial smartphones for fine-grained and real-time relative distance estimation. BAIR effectively aligns multiple sensory signals with different sampling rates via the improved Kalman filter technology. To mitigate IMU's integration errors, BAIR calculates the average velocity over a preceding period and uses this, alongside accumulated velocity data from the IMU, significantly improving distance prediction accuracy. We implemented our BAIR system on smartphones and conducted extensive experiments to evaluate its performance. Specifically, in static scenarios, BAIR achieves a mean average error (MAE) of 11 cm. In moving scenarios, the cumulative distribution function (CDF) values for 95%, 80%, and 50% are 31 cm, 13 cm, and 8 cm, respectively. The memory footprint of BAIR is 16.41 MB. We release a video demo on YouTube11https://youtu.be/7Fbmn4ALaI0.

Recently, Wi-Fi based sensing technology has been widely studied to provide more convenient services for humans. Although previous arts claim to achieve diverse fine-grained sensing using Wi-Fi signals, most of them assume that the data such as Channel State Information (CSI) used to achieve the sensing tasks can be sufficiently collected. However, in practical, due to the competitive nature of Wi-Fi and the frequent intermittent traffic, Wi-Fi based sensing applications often encounter problems of irregular intervals and insufficient sampling. Therefore, in this paper, we propose a Wi-Fi signal sparse recovery system (WiSR) that aims to recover sufficient and uniform sensing data from unevenly spaced and under-sampled CSI. Inspired by the success of image and audio restoration, we improve the Generative Adversarial Network (GAN) to recover Wi-Fi CSI. However, the direct application of GAN technologies for image and audio to CSI is not effective due to the difference in data representation. First, to avoid spectral impairments after conversion from time domain to frequency domain, we directly operate on the original time series CSI waveforms, thus being able to recover continuous channel variations from intermittent sparse samples. Second, to enhance the above recovery process, we utilize two novel denoising methods to obtain clean CSI, and introduce restrictions in the time and frequency domains to optimize low-level features and high-frequency information, respectively. Real-world experiments show that WiSR can accurately recover CSI, even at a rate of 10 packets per second. Through practical applications of gait recognition and gesture recognition, WiSR significantly improves accuracy compared to traditional linear interpolation and cubic interpolation.

In recent years, radio frequency (RF) signal-based sensing has garnered significant attention due to its ubiquity, with numerous applications emerging in areas such as target localization, material recognition, and health monitoring. However, current sensing models are often based on ray tracing, which, although computationally convenient, can become severely distorted when the target size is not much larger than the wavelength. Additionally, using signals with smaller wavelengths to mitigate this issue is not always feasible. Noting that RF signals are a form of electromagnetic waves, we have explored the development of field sensing models directly based on Maxwell's equations. These models can finely characterize phenomena such as diffraction and multiple scattering, thereby enhancing the upper limits of sensing system capabilities. Based on this approach, we have achieved integrated material recognition and imaging of centimeter-scale targets using WiFi signals. This work has been accepted for presentation at Ubicomp 2024.

Due to contactless and convenient experiences, WiFi-based device-free fingerprints localization technologies have extensively attracted research attention. However, they are studied based on an assumption that the user is stationary and face a major challenge in the presence of users motions. That is because users motions induced CSIWiFi variations results in inconsistent location fingerprints during training and prediction, leading to system ineffective. To solve this problem, in this paper, we propose freeLoc, which aims to free users motions (even unseen) while maintaining accurate localization. Specifically, we construct a domain adaptation network that defines different users and motions as different domains, and learns domain-independent representations to extract location fingerprints independent of users motions. Unfortunately, collecting sufficient amounts of WiFi data is difficult. To reduce the cost of labeling data and ensure the performance of domain adaptation network, we utilize adversarial autoencoder to build a data augmentation module to introduce data diversity. We deploy experiments in a real scenario, and the results show that only by labeling three motions of three users, we can achieve accurate localization (the nearest locations are about one meter away) for a total of 36 domains including 6 users and 6 motions. Compared to other existing technologies, freeLoc can improve location prediction accuracy by up to 35%.

The existing RF-based liquid identification systems usually rely on prior knowledge, such as pre-build database or the material and width information of the vessel. Furthermore, existing methods may not work in scenarios where the height of liquid is smaller than that of antenna. In this paper, we proposesLiqRay$^+$, a contactless system which can identify liquids in a fine-grained level without prior knowledge. To remove the effect of vessel, we build a dual-antenna model and craft a relative frequency response factor, exploring diversity of the permittivity in frequency domain. To eliminate the effect of different height, we devise the electric field distribution model at the receiving antenna, solving the unknown heights via spatio-differential model. Among eight different solvents,LiqRay$^+$can identify alcohol solutions with a concentration difference of 1% with 92.9% accuracy. Even if the liquid height is about 4 cm, which is fairly lower than that of most antennas’ heights, the accuracy is more than 85%.

Steganography is a technique that hides secret messages into a public multimedia object without raising suspicion from third parties. However, most existing works cannot provide good robustness against lossy JPEG compression while maintaining a relatively large embedding capacity. This paper presents an end-to-end robust steganography system based on the invertible neural network (INN). Instead of hiding in the spatial domain, our method directly hides secret messages into the discrete cosine transform (DCT) coefficients of the cover image, which significantly improves the robustness and anti-steganalysis security. A mutual information loss is first proposed to constrain the flow of information in INN. Besides, a two-way fusion module (TWFM) is implemented, utilizing spatial and DCT domain features as auxiliary information to facilitate message extraction. These two designs aid in recovering secret messages from the DCT coefficients losslessly. Experimental results demonstrate that our method yields significantly lower error rates than other existing hiding methods. For example, our method achieves reliable extraction with 0 error rate for 1 bit per pixel (bpp) embedding payload; and under the JPEG compression with quality factor QF=10, the error rate of our method is about 22% lower than the state-of-the-art robust image hiding methods, which demonstrates remarkable robustness against JPEG compression.

RFID technology has recently been exploited for not only identification but also fine-grained trajectory tracking and gesture recognition. While contact-based (a tag is attached to the target of interest) sensing has achieved promising results, contactless sensing still faces severe challenges such as low accuracy and inability to sense multiple targets simultaneously in proximity, restricting its applicability in real-world deployment. In this work, we present Tamera , a contactless RFID-based sensing system, which significantly improves the tracking accuracy, enables multi-commodity tracking, and even material and shopping behavior recognition. We successfully address multiple technical challenges, and design and implement our prototype on commodity RFID devices. We test the positioning accuracy of Tamera in a 5 m × 6 m laboratory. Tamera achieves a median error of 1.3 cm and 2.7 cm for contactless single- and multi-commodity tracking, respectively. In our laboratory, two shelves commonly found in the supermarket are arranged and the goods are placed on them. Tamera successfully localizes and identifies the material type (metal, plastic, paper, and glass) of the commodities on the shelf with an accuracy higher than 95%. Tamera successfully recognizes four shopping behaviors (taking commodity, replacing commodity, buying commodity, and invoking commodity) with an accuracy higher than 93%.

The existing RF-based liquid identification methods commonly require a training network of liquid or the container information, such as material and width. Moreover, status quo methods are inapplicable when the solution height is lower than that of the antenna, which is generally unknown either. This paper proposes LiqRay, an RF-based solution, retaining non-invasive and fine-grained liquid recognition abilities, thus can recognize unknown solutions without prior knowledge. In dealing with the unknown container material and width, we utilize a dual-antenna model and craft a relative frequency response factor, exploring diversity of the permittivity in frequency domain. In tackling the unknown heights of solution and antenna, we devise the electric field distribution model at the receiving antenna, solving the unknown heights via spatio-differential model. Among eight different solvents, LiqRay can identify alcohol solutions with a concentration difference of 1% with 94.92% accuracy. Nevertheless, LiqRay can obtain the relative frequency response factor with a relative error of 6.7% without being affected by the height of the solution. Even if it is merely 4 cm, this is fairly lower than that of most antennas' heights, since the operating frequency is around 2 GHz.

Due to nonlinearity and uncertainty of induction cooker system, it is difficult to establish an accurate mathematical model for control purpose. This paper proposes an appropriate fuzzy control strategy for controlling pan temperature of induction cooker system, without an accurate control model. In induction cooker system, the temperature feedback is from the glass countertop instead of the pan. Thus, the control precision needs be improved. The proposed fuzzy controller selects suitable output powers according to slew rates of the glass countertop temperature variation. It effectively reduces the temperature error between pan and glass countertop, which keeps the food in pan at a constant temperature. The proposed method is verified using the induction cooker with output power of about 1800W. Compared the fuzzy controller with PID controller in this system, a conclusion can be drawn that the fuzzy controller can provide a faster, more accurate and more stable temperature control.

Compressive sampling (CS) is a novel data collection and coding theory which allows us to recover sparse or compressible signals from a small set of measurements. This paper presents a new model for natural image recovery, in which the smooth l0norm and the approximate total-variation (TV) norm are adopted simultaneously. By using one-order gradient decrease, the speed of algorithm for this new model can be guaranteed. Experimental results demonstrate that the principle of the model is correct and the performance is as good as that based on TV model. The computing speed of the proposed method is two orders of magnitude faster than that of interior point method and two times faster than that of the Nesta optimization based on TV model.