MobiCom 2024 Papers

The majority of existing Wi-Fi backscatter systems transmit tag data at rates lower than 250 kbps, as the tag data is modulated at OFDM symbol level, allowing for demodulation using commercial Wi-Fi receivers. However, it is necessary to modulate tag data at OFDM sample level to satisfy the requirements for higher throughput. A comprehensive theoretical analysis and experimental investigation conducted in this paper demonstrates that demodulating sample-level modulated tag data using commercial Wi-Fi receivers is unattainable due to excessive computational overhead and demodulation errors. This is because the significant tag information dispersion, loss, and shuffling are caused by Wi-Fi physical layer operations. We conclude that the optimal position for demodulation is the time-domain IQ samples, which do not undergo any Wi-Fi physical layer operations and preserve the intact, ordered, and undispersed information of tag-modulated data, thereby minimizing complexity and maximizing accuracy.

Mobile devices contribute more than half of the world's web traffic, providing massive and diverse data for powering various federated learning (FL) applications. In order to avoid the communication bottleneck on the parameter server (PS) and accelerate the training of large-scale models on resource-constraint workers in edge computing (EC) system, we propose a novel split federated learning (SFL) framework, termed ParallelSFL. Concretely, we split an entire model into a bottom submodel and a top submodel, and divide participating workers into multiple clusters, each of which collaboratively performs the SFL training procedure and exchanges entire models with the PS. However, considering the statistical and system heterogeneity in edge systems, it is challenging to arrange suitable workers to specific clusters for efficient model training. To address these challenges, we carefully develop an effective clustering strategy by optimizing a utility function related to training efficiency and model accuracy. Specifically, ParallelSFL partitions workers into different clusters under the heterogeneity restrictions, thereby promoting model accuracy as well as training efficiency. Meanwhile, ParallelSFL assigns diverse and appropriate local updating frequencies for each cluster to further address system heterogeneity. Extensive experiments are conducted on a physical platform with 80 NVIDIA Jetson devices, and the experimental results show that ParallelSFL can reduce the traffic consumption by at least 21%, speed up the model training by at least 1.36X, and improve model accuracy by at least 5% in heterogeneous scenarios, compared to the baselines.

RF sensing has attracted a tremendous amount of attention and achieved promising progress in applications such as human gesture recognition and vital sign monitoring. This paper delves into sensing the moisture level of fabrics---an important metric for smart clothing, wound care, and textile manufacturing. We present TMSense, an innovative contact-free fabric moisture measurement system that leverages UWB signals for sensing. We introduce a set of signal processing methods to tackle the challenge of weak fabric reflections that can be easily overwhelmed by noise interference. Additionally, we adopt a model-driven approach to get rid of reliance on extensive datasets. By exploiting the changes in the dielectric properties induced by moisture in textile fabrics, we establish a theoretical model that bridges the characteristics of the RF signal with the moisture content. Based on this model, we successfully eliminate interfering factors such as target-device distance and target attributes through delicate signal processing and parameter calibration. Comprehensive experiments conducted under various conditions, including different materials, sample forms, and parameter settings, demonstrate an impressively low median error of 1.4% on textile moisture measurements, outperforming commodity moisture sensors on the market.

Bidirectional communication between BLE/FSK devices and WiFi access points (APs) combines the benefits of long battery life, low device cost, and ubiquitous Internet access. However, prior cross-technology communication (CTC) solutions require transmission mixers inside FSK chips, thus not applicable to newer ultra-low-power (ULP) BLE chips, which removes these mixers to conserve power. Furthermore, throughputs of prior CTC solutions are limited to 1Mbps.

Despite intensive research efforts in radio frequency noncontact sensing, capturing fine-grained geometric properties of objects, such as shape and size, remains an open problem using commodity WiFi devices. Prior attempts are incapable of characterizing object shape or size because they predominantly rely on weak signals reflected off objects in a very small number of directions. In this paper, motivated by the observation that the diffracted signals around an object between two WiFi devices carry the contour information of the object, we formulate the problem of reconstructing the 2D target profile and develop WiProfile, the first WiFi-based system that unlocks the diffraction effects for target profiling. We introduce a CSI-Profile model to characterize the relationship between the CSI measured at different target positions and the target profile in the diffraction zone. With suitable approximations, the inverse problem of deriving the target profile from CSI can be solved by the inverse Fresnel transform. To mitigate CSI measurement errors on commodity WiFi devices, we propose a novel antenna placement strategy. Comprehensive experiments demonstrate that WiProfile can accurately reconstruct profiles with median absolute errors of less than 1 cm under various conditions, and effectively estimate the profiles of everyday objects of diverse shapes, sizes, and materials. We believe this work opens up new directions for fine-grained target imaging using commodity WiFi devices.

Our perception of reality is under constant threat from ever-improving video manipulation techniques, including deep-fakes and generative AI. Therefore, proving authenticity of videos is increasingly important, especially in legal and news contexts. However, it is very challenging to prove it based on post-factum video content analysis.

In the era of ubiquitous short-video sharing, an overlooked yet significant privacy risk has arisen: the accidental disclosure of confidential information through reflective surfaces, such as mirrors, glass, or even polished metal. Such reflections can inadvertently disclose sensitive personal details to a broad audience without the awareness of content creators. Our examination of 100 top-rated TikTok short videos reveals that, on average, 37.2% of frames in each video feature identifiable reflective surfaces, posing potential privacy risks. In this work, we introduce a framework designed to scrutinize reflective privacy risks in glass-laden short videos. At the heart of the framework is the development of a reflection-specific neural radiance field, termed RP-NeRF, which enables reflection-aware ray tracing for precise extraction and reconstruction of the reflective scenes from the surfaces they appear on. A detailed field study on the framework indicates that the precision in detecting human presence and recognizing objects from the reconstructed reflective images reaches as high as 90.8% and 89.6%, respectively, even when dealing with a reflective surface that boasts 90% transparency and a mere 4% reflectance rate. These findings highlight the urgent need for greater awareness and advanced solutions to safeguard privacy in our digital age, especially in light of the significant impact of short-video sharing.

Fiducial marking is indispensable in mobile robots, including their pose calibration, contextual perception, and navigation. However, existing fiducial markers rely solely on vision-based perception which suffers such limitations as occlusion, energy overhead, and privacy leakage.

While Reconfigurable Intelligent Surfaces (RIS) have been shown to enhance OFDM communication performance, this paper unveils a potential security concern arising from widespread RIS deployment. Malicious actors could exploit vulnerabilities to hijack or deploy rogue RIS, transforming them from communication boosters into attackers. We present a novel attack that disrupts the critical orthogonality property of OFDM subcarriers, severely degrading communication performance. This attack is achieved by manipulating the RIS to generate frequency-shifted reflections/harmonics of the original OFDM signal. We also propose algorithms to simultaneously beamform the multiple RIS-generated frequency-shifted reflections towards selected targets. Extensive experiments conducted in indoor, outdoor, 3D, and office settings demonstrate that Hydra can achieve a 90% throughput reduction in targeted attack scenarios and a 43% throughput reduction in indiscriminate attack scenarios. Furthermore, we validated the effectiveness of our attacks on both the 802.11 protocol and the 5G NR protocol.

Recently, smart roadside infrastructure (SRI) has demonstrated the potential of achieving fully autonomous driving systems. To explore the potential of infrastructure-assisted autonomous driving, this paper presents the design and deployment of Soar, the first end-to-end SRI system specifically designed to support autonomous driving systems. Soar consists of both software and hardware components carefully designed to overcome various system and physical challenges. Soar can leverage the existing operational infrastructure like street lampposts for a lower barrier of adoption. Soar adopts a new communication architecture that comprises a bi-directional multi-hop I2I network and a downlink I2V broadcast service, which are designed based on off-the-shelf 802.11ac interfaces in an integrated manner. Soar also features a hierarchical DL task management framework to achieve desirable load balancing among nodes and enable them to collaborate efficiently to run multiple data-intensive autonomous driving applications. We deployed a total of 18 Soar nodes on existing lampposts on campus, which have been operational for over two years. Our real-world evaluation shows that Soar can support a diverse set of autonomous driving applications and achieve desirable real-time performance and high communication reliability. Our findings and experiences in this work offer key insights into the development and deployment of next-generation smart roadside infrastructure and autonomous driving systems.

Ensuring user privacy remains critical in mobile networks, particularly with the rise of connected devices and denser 5G infrastructure. Privacy concerns have persisted across 2G, 3G, and 4G/LTE networks. Recognizing these concerns, the 3rd Generation Partnership Project (3GPP) has made privacy enhancements in 5G Release 15. However, the extent of operator adoption remains unclear, especially as most networks operate in 5G Non Stand Alone (NSA) mode, relying on 4G Core Networks. This study provides the first qualitative and experimental comparison between 5G NSA and Stand Alone in real operator networks, focusing on privacy enhancements addressing top 8 pre-5G attacks based on recent academic literature. Also, it evaluates the privacy levels of OpenAirInterface, a leading open-source software for 5G, against real network deployments for the same attacks, revealing two new 5G privacy vulnerabilities, and underscoring the need for further studies and stricter standards.

Integrated sensing and communication (ISAC) is one of the usage scenarios for the sixth generation (6G) wireless networks. In this paper, we study cooperative ISAC in cell-free multiple-input multiple-output (MIMO) systems, where multiple MIMO access points (APs) collaboratively provide communication services and perform multistatic sensing. We formulate an optimization problem for the ISAC beamforming design, which maximizes the achievable sum-rate while guaranteeing the sensing signal-to-noise ratio (SNR) requirement and total power constraint. Learning-based techniques are regarded as a promising approach for addressing such a nonconvex optimization problem. By taking the topology of cell-free MIMO systems into consideration, we propose a heterogeneous graph neural network (GNN), namely SACGNN, for ISAC beamforming design. The proposed SACGNN framework models the cell-free MIMO system for cooperative ISAC as a heterogeneous graph and employs a transformer-based heterogeneous message passing scheme to capture the important information of sensing and communication channels and propagate the information through the graph network. Simulation results demonstrate the performance gain of the proposed SACGNN framework over a conventional null-space projection based scheme and a deep neural network (DNN)-based baseline scheme.

Legacy LoRa receiver adopts a superheterodyne architecture with a runtime power consumption of up to 100mW, resulting in its low-power promise can only be delivered in low duty-cycle mode. This paper presents Sisyphus as an ultra-low-power LoRa receiver, ensuring around-the-clock LoRa availability while extending battery life significantly. To achieve this, we propose a novel receiver design for passive coherent demodulation of LoRa. In this design, we creatively couple LoRa's down-conversion with de-chirping (dc2), leveraging the processing gain brought by chirp spread spectrum (CSS) modulation to boost communication range without the need for additional power supply. Moreover, we exploit the cyclical time-frequency feature intrinsic to LoRa for demodulation, and a low-power analog-digital signal processing circuit with negligible power is devised to replace the existing power-intensive sampling and costly digital computation. We prototype Sisyphus for proof-of-concept, and comprehensive experimental results demonstrate that Sisyphus can achieve significant power savings compared to legacy LoRa receiver while retaining the anti-interference ability of legacy LoRa. We envision that the design of Sisyphus can unlock the potential for broader applications of LoRa.

Constructing precise 3D maps is crucial for the development of future map-based systems such as self-driving and navigation. However, generating these maps in complex environments, such as multi-level parking garages or shopping malls, remains a formidable challenge. In this paper, we introduce a participatory sensing approach that delegates map-building tasks to map users, thereby enabling cost-effective and continuous data collection. The proposed method harnesses the collective efforts of users, facilitating the expansion and ongoing update of the maps as the environment evolves.

Low Earth Orbit (LEO) satellite mega-constellations promise ubiquitous network services to "unconnected" users. But their upcoming global routing for Earth will be unstable due to exhaustive topology updates between satellites and Earth, inside an orbital shell, and across heterogeneous orbital shells. In real LEO networks, these multi-dimensional dynamics are interleaved and complicated by chaotic orbital maneuvers and random failures. They are less predictable than most satellite routing proposals expect and threaten these proposals' availability, efficiency, or resiliency at scale.

The Open RAN architecture, with disaggregated and virtualized RAN functions communicating over standardized interfaces, promises a diversified and multi-vendor RAN ecosystem. However, these same features contribute to increased operational complexity, making it highly challenging to troubleshoot RAN related performance issues and failures. Tackling this challenge requires a dependable, explainable anomaly detection method that Open RAN is currently lacking. To address this problem, we introduce SpotLight, a tailored system archtecture with a distributed deep generative modeling based method running across the edge and cloud. SpotLight takes in a diverse, fine grained stream of metrics from the RAN and the platform, to continually detect and localize anomalies. It introduces a novel multi-stage generative model to detect potential anomalies at the edge using a light-weight algorithm, followed by anomaly confirmation and an explain-ability phase at the cloud, that helps identify the minimal set of KPIs that caused the anomaly. We evaluate SpotLight using the metrics collected from an enterprise-scale 5G Open RAN deployment in an indoor office building. Our results show that compared to a range of baseline methods, SpotLight yields significant gains in accuracy (13% higher F1 score), explain-ability (2.3 -- 4X reduction in the number of reported KPIs) and efficiency (4 -- 7X bandwidth reduction).

Underwater positioning plays a key role in many underwater operations. This paper presents the design, implementation, and evaluation of UWBeacon, a centimeter-level visible light-based underwater positioning system. UWBeacon consists of LED beacons as the light signal transmitter and a camera-based receiver as the target. To address unique challenges in underwater environment such as limited visibility and strong ambient interference, we exploit a novel design that utilizes polarized lights of different colors with different polarization angles for background subtraction. UWBeacon is implemented with commercial-off-the-shelf LEDs and cameras. Comprehensive experiments conducted in various real underwater environments show that UWBeacon can achieve a mean positioning error below 6 cm and an orientation error below 1.5° at a distance of 10 meters.

Modern speech understanding (SU) runs a sophisticated pipeline: ingesting streaming voice input, the pipeline executes encoder-decoder based deep neural networks repeatedly; by doing so, the pipeline generates tentative outputs (called hypotheses), and periodically scores the hypotheses.

While video recommendation has been studied extensively in regular PC and smartphone settings, such a topic has been rarely discussed in the virtual reality (VR) context so far. On the other hand, as the popularity of VR videos continues to soar, its recommendation will play a crucial part in providing suggestions and guiding users through a deluge of available content. Given this unmet need, in this work, we present Bere, a video recommender system tailored for VR. Our approach leverages viewers' behavioral responses as they engage with VR videos to infer their preferences and thus make future recommendations. We integrate these new behavioral user-video interaction measures into the mainstream recommendation framework and renovate the graph learning-based paradigm to accommodate the new changes. The recommender system is further empowered with a novel domain adaptation approach named CMCCDA to address the data scarcity problem for model training. We also develop an energy-efficient adaptive encoding scheme to reduce the energy consumption on the VR device. We collect a behavioral dataset for video recommendation in VR and demonstrate through extensive evaluation that Bere significantly outperforms state-of-the-art schemes by up to 68.0% in precision and up to 28.8% in ranking quality.