Structured data is well handled by gradient-boosted decision trees (GBDT), which are usually trained on vertically partitioned features across mutually distrustful parties. High speed and interpretability make GBDTs popular in finance and healthcare, where neural networks may fall short. Enabling secure computation for GBDTs poses unique challenges, requiring secure record alignment for comparison. Relying on private set intersection (PSI) is a de facto approach. Mistaking PSI for a safety measure actually exposes which record identifiers (IDs) are shared between the datasets. Although circuit-PSI could help, it is costly for generic uses. New ideas are needed to efficiently train in a "dark forest". Aiming to hide the IDs, we initiate the study of anonymous GBDT training on split data held by two parties. Dual circuit-PSI in our design lets the parties alternate as receiver to run pick-then-sum over local features. Via oblivious programmable pseudorandom functions, we propagate circuit-PSI outputs as shared state across runs. Avoiding universal alignment, we resolve the neglected dilemma that ID hiding incurs a cost that scales with domain size. Next, we halve the cost of ciphertext packing used to convert single-instruction multiple-data homomorphic encryption from (ring) learning with errors in prior secure GBDT (Usenix Security' 23) and related secure machine-learning computations. Comparative experiments show our protocol remains competitive with leaky approaches in efficiency. Enabling ID-hiding aggregation, our techniques can extend to other vertically partitioned analytics.

3D scene understanding is an important task, and there has been a recent surge of research interest in aligning 3D representations of point clouds with text to empower embodied AI. However, due to the lack of comprehensive 3D benchmarks, the capabilities of 3D models in real-world scenes, particularly those that are challenging with subtly distinguished objects, remain insufficiently investigated. To facilitate a more thorough evaluation of 3D models' capabilities, we propose a scheme, ObjVariantEnsemble, to systematically introduce more scenes with specified object classes, colors, shapes, quantities, and spatial relationships to meet model evaluation needs. More importantly, we intentionally construct scenes with similar objects to a certain degree and design an LLM-VLM-cooperated annotator to capture key distinctions as annotations. The resultant benchmark can better challenge 3D models, reveal their shortcomings in understanding, and potentially aid in the further development of 3D models.

In an era dominated by big data and machine learning, establishing valuable data collaboration has never been more critical. However, such collaborations must operate under regulatory and legal constraints. Two-party Privacy-Preserving Record Linkage (PPRL) emerges to assess the potential collaboration value and also ensure the privacy and security of the involved data. Nevertheless, the substantial computational and communication overheads associated with PPRL hinder its practical adoption in data markets with numerous potential collaborators. Therefore, we present the Screening-then-Linkage framework, which incorporates a lightweight Screening phase prior to the resource-intensive PPRL phase, i.e., PPRS, to mitigate the scalability issue of PPRL. We propose a circuit-PSI-based system, named Appraisal to realize a secure, effective, and efficient PPRS. To reconcile the approximate matching and/or schema-aware setting required in PPRS with the limitations of the circuit-PSI supporting only symmetric functions, we propose a more communication-efficient secure permutation, i.e., Oblivious Attribute/Feature Alignment protocol tailored for PPRS. This protocol supports a broader range of comparison functions and significantly improves efficiency, i.e., reducing communication costs by a factor of 14 compared to the conventional protocol. Our rigorous analysis and comprehensive empirical evaluations demonstrate the security, effectiveness, and efficiency of Appraisal. Appraisal can accommodate up to 850x more records than the SOTA PPRS system, SFour, within the same constraints. Moreover, it is 165x faster than SOTA PPRL, indicating the Screening-then-Linkage framework substantially decreases the computation time required to identify the most valuable collaborators from a large pool of candidates.

In this poster, we present our efforts to enable personalized LLM services in smart environments through persistent context memory. By systematically organizing long-term sensor logs into a structured format that captures user's behavioral patterns/preference in the environment, our solution MemAura empowers LLMs to better understand user intents and deliver tailored services. With its effective memory management, MemAura holds promise to make smart living spaces more efficient, context-aware and user-centric.

There is a growing trend in deploying DNNs on tiny microcontroller (MCUs) to provide inference capabilities in the IoT. While prior research has explored many lightweight techniques to compress DNN models, achieving overall efficiency in model inference requires not only model optimization but also careful system resource utilization for execution. Existing studies primarily leverage arithmetic logic units (ALUs) for integer-only computations on a single CPU core. Floating-point units (FPU) and multi-core capabilities available in many existing MCUs remain underutilized. To fill this gap, we propose AdaptQNet, a novel MCU neural network system that can determine the optimal precision assignment for different layers of a DNN model. AdaptQNet models the latency of various operators in DNN models across different precisions on heterogeneous processing units. This facilitates the discovery of models that utilize FPU and multi-core capabilities to enhance capacity while adhering to stringent memory constraints. Our implementation and experiments demonstrate that AdaptQNet enables the deployment of models with better accuracy-efficiency trade-off on MCUs.

The increasing use of camera streams on mobile systems has raised significant privacy concerns due to unauthorized visual data access by applications. Existing solutions either burden users with excessive interaction or lack semantic understanding of contextual privacy norms. This paper introduces PrivacyAgent, a novel visual privacy protection framework leveraging multimodal large language models (LLMs) to enable context-aware and fine-grained privacy control on mobile systems. PrivacyAgent intercepts camera streams via a virtualized I/O layer and restricts untrusted apps to privacy-compliant content with minimal user overhead.

In this paper, we propose an innovative method for fingerprinting USB peripherals. While USB technology has made significant progress in data transfer efficiency, by default, the USB host device trusts any connected peripherals. This ignorance has led to several security risks in practice. Existing protection practices mitigate these risks by verifying the specific type or even the identity of the peripheral before data transfer. However, these methods often depend on expensive and specialized hardware, such as software-defined radios, limiting their applicability in everyday scenarios.To address this issue, we propose profiling the electromagnetic radiation (EMR) generated by a USB peripheral as its unique identifier. Our method utilizes a low-cost backscatter unit and the host’s WiFi network card for the profiling and verification process. The backscatter unit is integrated into a USB hub. It shifts the USB EMR signal into the frequency domain within the WiFi sensing range, and transforms EMR into an interpretable form suitable for feature extraction. Subsequently, we employ a hybrid classification method, combining heuristic and neural network clues, to recognize the identity of the EMR’s source. We evaluate the effectiveness of our method on an extensive dataset collected from USB peripherals of various types and vendors.

Accurate channel estimation is crucial for the performance gains of massive multiple-input multiple-output (mMIMO) technologies. However, it is bandwidth-unfriendly to estimate large channel matrix frequently to combat the time-varying wireless channel. Deep learning-based channel prediction has emerged to exploit the temporal relationships between historical and future channels to address the bandwidth-accuracy trade-off. Existing methods with convolutional or recurrent neural networks suffer from their intrinsic limitations, including restricted receptive fields and propagation errors. Therefore, we propose a Transformer-based model, CS3T-UNet tailored for mMIMO channel prediction. Specifically, we combine the cross-shaped spatial attention with a group-wise temporal attention scheme to capture the dependencies across spatial and temporal domains, respectively, and introduce the shortcut paths to well-aggregate multi-resolution representations. Thus, CS3T-UNet can globally capture the complex spatial-temporal relationship and predict multiple steps in parallel, which can meet the requirement of channel coherence time. Extensive experiments demonstrate that the prediction performance of CS3T-UNet surpasses the best baseline by at most 6.86 dB with a smaller computation cost on two channel conditions.

Remote conference applications are increasingly widely used, but currently, their improper data encryption methods, proprietary implementations, and dial-in access cause concerns about privacy breaches. As such, there is a need for trustworthy and secure solutions for these production tools. In this paper, we present mTunnel, a transparent software layer in the host system for securing conference applications without sacrificing the key functionalities and convenience. The basic idea of mTunnel is to encrypt sensitive data, such as audio, video, text, etc., before it is obtained by untrusted application clients. mTunnel leverages the audio and video streaming capabilities of the conference applications to tunnel the encrypted content to the remote end. mTunnel involves a software framework to accommodate the media interception and representation through I/O virtualization based on virtual drivers. Moreover, mTunnel supports complete E2EE group conversations even in a mixed IP and public switched telephone network (PSTN). We implement mTunnel and evaluate it with several commercial products. Results show its feasibility and overhead.

Head tracking is a technique that allows for the measurement and analysis of human focus and attention, thus enhancing the experience of human-computer interaction (HCI). Nevertheless, current solutions relying on vision and motion sensors exhibit limitations in accuracy, user-friendliness, and compatibility with the majority of commercial off-the-shelf (COTS) devices. To overcome these limitations, we present, an earphone-based system that achieves head tracking exclusively through acoustic signals. employs acoustic sensing to measure the movement of a pair of earphones, subsequently enabling precise head tracking. In particular, a pair of speakers generates a periodically fluctuating sound field, which the user’s two earphones detect. By assessing the distance and angle alterations between the earphones and speakers, we propose a model to determine the user’s head movement and orientation. Our evaluation results indicate a high degree of accuracy in both head movement tracking, with an average tracking error of 2.98 cm, and head orientation tracking, with an average error of 1.83 degrees. Furthermore, in a deployed exhibition scenario, we attained an accuracy of 89.2% in estimating the user’s focus direction.

The massive MIMO gain for wireless communication has been greatly hindered by the feedback overhead of channel state information (CSI) growing linearly with the number of antennas. Recent efforts leverage the DNN-based encoder-decoder framework to exploit correlations within the CSI matrix for better CSI compression. However, existing works have not fully exploited the unique features of CSI, resulting in an unsatisfactory performance under high compression ratios and sensitivity to multipath effects. Instead of treating CSI as common 2D matrices like images, we reveal the intrinsic stripe-based correlation across the CSI matrix. Driven by this insight, we propose CSI-StripeFormer, a stripe-aware encoder-decoder framework to exploit the unique stripe feature for better CSI compression. We design a lightweight encoder with asymmetric convolution kernels to capture various shape features. We further incorporate novel designs tailored for stripe features, including a novel hierarchical Transformer backbone in the decoder and a hybrid attention mechanism to extract and fuse correlations in angular and delay domains. Our evaluation results show that our system achieves an over 7dB channel reconstruction gain under a high compression ratio of 64 in multipath-rich scenarios, significantly superior to current state-of-the-art approaches. This gain can be further improved to 17dB given the extended embedded dimension of our backbone.

Multi-user indoor localization is considered to be one of the most useful wireless applications. Low latency and high robustness to dynamic interference from surrounding people are essential requirements for multi-user localization. However, state-of-the-art (SOTA) indoor localization systems cannot satisfy both requirements at the same time. In this paper, we propose RIScan, a Reconfigurable Intelligent Surface (RIS)-aided localization system that can achieve both low latency and high reliability. We leverage RIS to perform Wi-Fi beam scanning so all clients can figure out their direction in a single scan. However, compared with traditional AP-based systems, the introduction of RIS creates a more complicated signal superposition at the receiver, preventing clients from directly obtaining target beams for direction derivation and localization. To overcome this challenge, we fully utilize the reconfigurability of RIS to endow target beams with distinguishing features, so that RIScan can extract stable and accurate direction information from complex and dynamic environments. RIScan is implemented in the real system with our own developed 16 × 16 RIS prototype and COTS Wi-Fi devices. Extensive experiments show that RIScan achieves a median localization error of 47cm and 71cm in static and dynamic environments with only two RIS anchors. Compared to the SOTA methods, RIScan reduces the localization latency by more than an order of magnitude.

The emerging reconfigurable intelligent surface (RIS) technique introduces novel threats to wireless sensing owing to its channel customization ability. Unlike active radios, the RIS's interference behaves akin to natural reflections, exhibiting a higher level of stealthiness and difficulty in detection. However, the majority of current RIS-based attacks lack generalizability to real-world scenarios, as they assume complete coverage of the RIS over objects and develop their techniques within electromagnetic-controlled environments such as an anechoic chamber. To bridge this gap, we present RIStealth, a practical and covert attack that leverages RIS technology to render a moving individual undetectable by WiFi-based intrusion detection systems in real-life scenarios. RIStealth integrates the strengths of both motion reduction and threshold lifting strategies to address challenges of limited RIS affordability, constrained cooperation in adversary settings, and complex and unpredictable environments. Through real-world evaluations conducted with our RIS prototype, we demonstrate that RIStealth effectively reduces the victim's intrusion detection rate from 95.1% to 16.4%. Our findings shed light on the practical threats posed by RIS, thereby encouraging further countermeasure development.

Graph alignment, also known as network alignment, has many applications in data mining tasks. It aims to find the node correspondence across disjoint graphs. With recent representation learning advancements, embedding-based graph alignment has become a hot topic. Existing embedding-based methods focus either on structural proximity across graphs or on the positional proximity within a single graph. However, only considering the structural similarity will make the position relation of nodes not clear enough, which makes it easy to misalign the nodes close in distance, while only considering the position proximity of a single graph will make the node embeddings from different graphs in different subspaces. To mitigate this issue, we propose a novel model CEGA forCross-graphEmbedding-basedGraphAlignment, which can generate node embeddings to reflect structural proximity and positional proximity simultaneously. Meanwhile, we make the proximity trainable thus it can be learned to best suit the alignment task at hand automatically. We show that CEGA outperforms existing graph alignment methods in accuracy under unsupervised scenarios through extensive experiments on public benchmarks.

Machine Learning as a Service is a promising service for individuals and companies who would like to delegate model training to third parties. The customers desire proof of the integrity of the model training to prevent potential backdoor attacks launched by the server, while the server desires to prove the integrity without revealing their intellectual assets, hyper-parameters of the training scheme. Zero-knowledge proof, a cryptographic tool can theoretically satisfy the above demand, but is still practically infeasible due to the inefficiency of proving. Thus, we propose zkMLaaS, a privacy-preserving and verifiable scheme for efficient training proof generation in the MLaaS scenario. zkMLaaS features a two-round challenge-response pro-tocol equipped with the random sampling. This greatly reduces the time cost of proof generation and ensures the integrity of training procedure simultaneously. We analyze the security of zkMLaaS and conduct comprehensive evaluation which shows it saves around$273\times$times compared with naive scheme.

On-boarding new devices into an existing SDN network is a pain for network operations (NetOps) teams, because much expert effort is required to bridge the gap between the configuration models of the new devices and the unified data model in the SDN controller. In this work, we present an assistant framework NAssim, to help NetOps accelerate the process of assimilating a new device into a SDN network. Our solution features a unified parser framework to parse diverse device user manuals into preliminary configuration models, a rigorous validator that confirm the correctness of the models via formal syntax analysis, model hierarchy validation and empirical data validation, and a deep-learning-based mapping algorithm that uses state-of-the-art neural language processing techniques to produce human-comprehensible recommended mapping between the validated configuration model and the one in the SDN controller. In all, NAssim liberates the NetOps from most tedious tasks by learning directly from devices' manuals to produce data models which are comprehensible by both the SDN controller and human experts. Our evaluation shows, NAssim can accelerate the assimilation process by 9.1x. In this process, we also identify and correct 243 errors in four mainstream vendors' device manuals, and release a validated and expert-curated dataset of parsed manual corpus for future research.

With the arrival of the Internet of Things era, IoT devices and the services built on them make our lives more convenient and also raise public concerns on their vulnerability to attacks. Recent literature advocates physical-layer solutions to help IoT devices detect attacks instead of using sophisticated cryptographic methods. However, there is still no satisfying solutions for IoT devices with a single antenna and sparse traffic. Thus, we introduce SecurePilot to fill this gap. SecurePilot is an unsupervised and plug-and-play solution which works without an attacker’s knowledge in advance. It leverages the strengths of two orthogonal physical-layer information, propagation signatures and device signatures embedded in pilot signals to enable effective attack detection. It could work on single-antenna IoT devices with sparse traffic and also work compatibly with communication protocols. The experimental results show that SecurePilot can successfully detect 99.6% of attacks, triggering false alarms on 3.1% of legitimate traffic in a typical office environment.

Reputation/trust-based blockchain systems have attracted considerable research interests for better integrating Internet of Things with blockchain in terms of throughput, scalability, energy efficiency, and incentive aspects. However, most existing works only consider static adversaries. Hence, they are vulnerable to slowly adaptive attackers, who can target validators with high reputation value to severely degrade the system performance. Therefore, we introduce$\textsf{zkRep}$, a privacy-preserving scheme tailored for reputation-based blockchains. Our basic idea is to hide both the identity and reputation of the validators by periodically changing the identity and reputation commitments (i.e., aliases), which makes it much more difficult for slowly adaptive attackers to identify validators with high reputation value. To realize this idea, we utilize privacy-preserving Pedersen-commitment-based reputation updating and leader election schemes that operate on concealed reputations within an epoch. We also introduce a privacy-preserving identity update protocol that changes the identity and time-window-based cumulative reputation commitments during each epoch transition. We have implemented and evaluated$\textsf{zkRep}$on the Amazon Web Service. The experimental results and analysis show that$\textsf{zkRep}$achieves great privacy-preserving features against slowly adaptive attacks with little overhead.

Smart speakers play an important role in smart home envision. Active acoustic sensing can enable convenient gesture interaction on smart speakers to complement voice interaction in mandatory quiet scenarios and address privacy concerns. However, existing solutions did not consider the impact of the widely adopted uniform circular geometry of commercial smart speakers on gesture tracking. To fill this gap, we propose SparseTrack to achieve fine-grained multi-user device-free gesture tracking on commercial smart speakers. We cast gesture tracking to sparse recovery intuition to address signal coherence issue on uniform circular mic-array. We then synthesize wideband measurement to eliminate spatial ambiguity caused by the insufficient spatial sampling rate of today’s smart speakers in the ultrasonic frequency band. We further design a robust trace extraction approach and properly handle the impact of the doppler effect on gesture tracking. We implement SparseTrack on COTS circular mic-array and conduct extensive evaluations. The results show that our system can simultaneously track up to 4 users’ gestures with a mean tracking error of 2.66 cm.

In today's blockchain system, designing a secure and high throughput blockchain on par with a centralized payment system is a difficult task. Sharding is one of the most worthwhile emerging technologies for improving the system throughput while maintain high-security level. However, previous sharding-related designs have two main limitations. First, the security and throughput of their random-based sharding system are not high enough as they did not leverage the heterogeneity among validators. Second, to design an incentive mechanism that promotes cooperation could incur a huge overhead on their system. In this article, we propose RepChain, a reputation-based secure and fast blockchain system via sharding, which also provides high incentive to stimulate node cooperation. RepChain utilizes reputation to explicitly characterize the heterogeneity among the validators and lay the foundation for the incentive mechanism. We propose a new double-chain architecture-a transaction chain and a reputation chain. For the transaction chain, an efficient Raft-based synchronous consensus has been presented. For the reputation chain, the synchronous Byzantine fault tolerance consensus that combines collective signing has been utilized to prevent the attack on both reputation score and the related transaction blocks. It supports a high throughput transaction chain with moderate generation speed. Moreover, we propose a reputation-based sharding and leader selection scheme. To analyze the security of RepChain, we propose a recursive formula to calculate the epoch security within only $\mathcal {O}(km^{2})$ time. Furthermore, we implement and evaluate RepChain on the Amazon Web Service platform. The results show our solution can enhance both throughout and security level of the existing sharding-based blockchain system.

Deep neural networks (DNNs)-powered Electrocardiogram (ECG) diagnosis systems recently achieve promising progress to take over tedious examinations by cardiologists. However, their vulnerability to adversarial attacks still lack comprehensive investigation. The existing attacks in image domain could not be directly applicable due to the distinct properties of ECGs in visualization and dynamic properties. Thus, this paper takes a step to thoroughly explore adversarial attacks on the DNN-powered ECG diagnosis system. We analyze the properties of ECGs to design effective attacks schemes under two attacks models respectively. Our results demonstrate the blind spots of DNN-powered diagnosis systems under adversarial attacks, which calls attention to adequate countermeasures.

Face authentication systems have gained widespread popularity because of their user-friendly usage and increasing recognition accuracy. Unfortunately, the boom in mobile social networks has bought with it media-based facial forgery; a critical threat where an adversary forges or replays the victim's photograph/video to fool the system. In this article, we propose EchoFace, an effective and robust liveness detection system to enhance face authentication in defending against media-based attacks, which works with today's smartphones/smartwatches without any hardware modification. EchoFace uses active acoustic sensing to differentiate the uneven stereostructure of the face and the flat forged media. Our proposed scheme effectively extracts the desired reflection profiles from the target. Moreover, we propose effective similarity measurements of reflection profiles to distinguish live users from forged media, which works robustly under various environmental conditions. EchoFace only requires low cost and universally equipped acoustic sensors without human intervention for liveness detection, which can be easily deployed in a variety of application scenarios. We implement EchoFace on commercial smartphones, and experiment results show that EchoFace achieves an average detection accuracy higher than 96% and false alarm rate lower than 4% across various media attacks and different levels of background noise. This shows its great potential to enhance the security of widely deployed face authentication systems in real scenarios.

In the past decade, with the rapid development of wireless communication and sensor technology, ubiquitous smartphones equipped with increasingly rich sensors have more powerful computing and sensing abilities. Thus, mobile crowdsensing has received extensive attentions from both industry and academia. Recently, plenty of mobile crowdsensing applications come forth, such as indoor positioning, environment monitoring, and transportation. However, most existing mobile crowdsensing systems lack vast user bases and thus urgently need appropriate incentive mechanisms to attract mobile users to guarantee the service quality. In this paper, we propose to incorporate sensing platform and social network applications, which already have large user bases to build a three-layer network model. Thus, we can publicize the sensing platform promptly in large scale and provide long-term guarantee of data sources. Based on a three-layer network model, we design incentive mechanisms for both intermediaries and the crowdsensing platform and provide a solution to cope with the problem of user overlapping among intermediaries. We theoretically prove the properties of our proposed incentive mechanisms, including incentive compatibility, individual rationality, and efficiency. Furthermore, we evaluate our incentive mechanisms by extensive simulations. Evaluation results validate the effectiveness and efficiency of our proposed mechanisms.

This paper presents DriveSafe Inspector, a fine-grained driver hand position monitoring system, which continuously detects a driver's hand position on the steering wheel. The steering wheel is divided into twelve 30° sectors like a clock. Our system can be applied on off-the-shelf hardware and works without extra modification to vehicles. In our system, sensor readings from both a wearable and its paired smartphone are fused to infer hand posture and turning angle between static holding states. With both static holding and dynamic turning information, our system achieves fine-grained hand position prediction in the presence of diverse road conditions and inter-individual differences. The on-road evaluation shows that our system can achieve an average 91.59% hand position detection accuracy with only static information, and can be further improved to 94.63% accuracy combined with dynamic turning information.

Nowadays, wrist wearables are widely used in applications containing privacy sensitive data. Protecting these data from unauthorized access is highly demanded. However, Personal Identification Number(PIN)-based certificate is not convenient to use for wearables are small in size and have no/very limited input ways. Like signing on paper, signing a pattern in the air and authenticate the user according to the trait from motion sensors is a potential solution for the authentication needs on wearables. However, we observe the successful authentication rate of such approaches in existing designs is not high enough under different postures. We identify the problem is caused by the residue gravity acceleration in motion sensors which introduces inaccuracy in capturing signing dynamic. To this end, this paper proposes a robust gesture-based authentication method that can be used under different postures. Our key idea to use and only use the gyroscope sensor to capture the signing dynamics, so that to get rid of the impact from gravity. To realize this idea, we propose an accurate wrist dynamic capturing method with gyroscope sensor, through which behavior traits are extracted from both time and shape domain. Finally, by comparing the Dynamic Time Warping distance between the real time input and the traits template, signing user can be identified. We implement our method on android smart watch. Our evaluation involves 11 users and lasts for more than one month. The results show that our method can achieve 90.1% balanced accuracy and have stable performance under different postures.

In the past decade, with the rapid development of wireless communication and sensor technology, ubiquitous smartphones equipped with increasingly rich sensors have more powerful computing and sensing abilities. Thus, mobile crowdsensing has received extensive attentions from both industry and academia. Recently, plenty of mobile crowdsensing applications come forth, such as indoor positioning, environment monitoring, transportation, and so on. However, most existing mobile crowdsensing systems lack of vast user bases, and thus urgently need appropriate incentive mechanisms to attract mobile users to guarantee the service quality. In this paper, we propose to incorporate sensing platform and social network applications, which already have large user bases to build a three-layer network model. Thus, we can publicize the sensing platform promptly in large scale, and provide longterm guarantee of data sources. Based on a three-layer network model, we design incentive mechanisms for both intermediaries and the crowdsensing platform, and provide a solution to cope with the problem of user overlapping among intermediaries. We indicate the properties of our proposed incentive mechanisms, including incentive compatibility, individual rationality, and efficiency.