Accurate multi-organ segmentation in Computed Tomography (CT) images is critical for computer-aided diagnosis systems. However, existing supervised methods heavily rely on costly, high-quality labeled data. To address this, we propose a label-efficient segmentation method for abdominal organs in CT images, leveraging knowledge transfer from a pre-trained diffusion model. Specifically, we pre-train a denoising diffusion model on 207,029 unlabeled 2D CT slices to capture anatomical patterns, which is then fine-tuned on limited labeled data for abdominal organ segmentation. During fine-tuning, two strate-gies-linear probing and decoder fine-tuning-are employed to adapt the model for segmentation while preserving learned representations. Quantitative results demonstrate that the pre-trained diffusion model can generate diverse and realistic$256 \times 256$CT images (FID: 11.32, sFID: 46.93, F1-score: 73.1%). Moreover, our method achieves competitive performance on the FLARE 2022 dataset for organ segmentation, particularly excelling in limited labeled data scenarios. With only 10% and 1% labeled data, our method achieves DSCs of 78.51% and 71.56% on 13 abdominal organs, respectively. Remarkably, with only four labeled 2D slices, our method still achieves a DSC of 51.81%, highlighting the efficacy of our method in alleviating the reliance of supervised learning on large-scale labeled data.

Federated Learning (FL) enables decentralized model training while preserving data privacy. Despite its benefits, FL faces challenges with non-identically distributed (non-IID) data, especially in long-tailed scenarios with imbalanced class samples. Momentum-based FL methods, often used to accelerate FL convergence, struggle with these distributions, resulting in biased models and making FL hard to converge. To understand this challenge, we conduct extensive investigations into this phenomenon, accompanied by a layer-wise analysis of neural network behavior. Based on these insights, we propose FedWCM, a method that dynamically adjusts momentum using global and per-round data to correct directional biases introduced by long-tailed distributions. Extensive experiments show that FedWCM resolves non-convergence issues and outperforms existing methods, enhancing FL’s efficiency and effectiveness in handling client heterogeneity and data imbalance.

Global aging necessitates tele-monitoring systems to provide real-time tracking and timely assistance for older adults living independently. While pervasive wireless devices (e.g., CSI, IMU, UWB) enable cost-effective, non-intrusive monitoring, existing systems lack flexibility, limiting their adaptability to different environments. In this work, we posit that the motion dynamics of human movement are invariant across sensing modalities, inspiring the design of HARMONY—a privacy-preserving, sensor-agnostic system that supports multi-modal inputs and diverse tele-monitoring tasks. HARMONY incorporates Modality-agnostic Data Processing to uniformly encrypt multi-modal signals and Task-specific Activity Recognition for seamless tasks adaptation. A novel Encrypted-processing Engine then significantly accelerates computations on encrypted data by optimizing matrix and convolution operations. Evaluations across five different sensing modalities show that HARMONY consistently achieves high accuracy while delivering 3.5 × to 130 × speedups over state-of-the-art baselines. Our results demonstrate that HARMONY is a practical, scalable, and privacy-centric prototype for next-generation remote healthcare.

In federated learning (FL), models must \emph{converge quickly} under tight communication budgets while \emph{generalizing} across non-IID client distributions. These twin requirements have naturally led to two widely used techniques: client/server \emph{momentum} to accelerate progress, and \emph{sharpness-aware minimization} (SAM) to prefer flat solutions. However, simply combining momentum and SAM leaves two structural issues unresolved in non-IID FL. We identify and formalize two failure modes: \emph{local–global curvature misalignment} (local SAM directions need not reflect the global loss geometry) and \emph{momentum-echo oscillation} (late-stage instability caused by accumulated momentum). To our knowledge, these failure modes have not been jointly articulated and addressed in the FL literature. We propose \textbf{FedWMSAM} to address both failure modes. First, we construct a momentum-guided global perturbation from server-aggregated momentum to align clients' SAM directions with the global descent geometry, enabling a \emph{single-backprop} SAM approximation that preserves efficiency. Second, we couple momentum and SAM via a cosine-similarity adaptive rule, yielding an early-momentum, late-SAM two-phase training schedule. We provide a non-IID convergence bound that \emph{explicitly models the perturbation-induced variance} $\sigma_\rho^2=\sigma^2+(L\rho)^2$ and its dependence on $(S,K,R,N)$ on the theory side. We conduct extensive experiments on multiple datasets and model architectures, and the results validate the effectiveness, adaptability, and robustness of our method, demonstrating its superiority in addressing the optimization challenges of Federated Learning. Our code is available at \url{https://github.com/Li-Tian-Le/NeurlPS_FedWMSAM}.

The rise of spy cameras has become a global concern regarding personal privacy. Unfortunately, affordable and easily accessible detection methods are lacking, leaving individuals vulnerable to these intrusive devices. Currently, most research focuses on detecting cameras based on their wireless transmission. Some of the latest research can detect other types of emissions from cameras but require costly equipment like universal software radio peripherals (USRP) or thermal cameras. However, this paper introduces a ubiquitous, cost-effective (less than $2) system, CamFirm, to identify and locate spy cameras regardless of their transmission mode. CamFirm only requires wired headphones and a Frequency Modulation (FM) module. Despite limitations of the FM module, such as narrow bandwidth, distortion, and vulnerability to interference, CamFirm remains capable of accurately identifying camera signals and pinpointing their locations. By utilizing the harmonics of digital circuits for signal screening, CamFirm can detect the photosensitive sensor signature of cameras and ascertain its direction through the attenuation of electromagnetic waves by the human body. The camera’s position can be further confirmed by activating a phone’s flashlight. Through extensive testing on 18 different cameras, CamFirm can achieve a median detection distance of 2.6 meters, effectively guiding users in locating the hidden camera.

Vehicular Ad-hoc Networks (VANETs) face significant security and privacy challenges in modern intelligent transportation systems. This paper analyzes vulnerabilities in Al-Shareeda et al.'s vehicle authentication protocol (doi: 10.1109/TDSC.2025.3553868) and proposes an enhanced ECC-based scheme using short-lived pseudonymous certificates. We identify two critical weaknesses in Al-Shareeda et al.'s protocol—a desynchronization attack causing potential denial-of-service and an identity linking attack compromising vehicle privacy. Our protocol establishes mutual authentication between vehicles and roadside units, ensuring message integrity, anonymity, and perfect forward secrecy. Unlike existing approaches, it eliminates the need for online third-party authenticators. Formal security proofs demonstrate that the scheme's security is reducible to the hardness of the ECDLP and CDH problems. Performance analysis shows our approach achieves an optimal security-efficiency balance with competitive communication overhead (4608 bits) and computation costs (5.02 ms) compared to state-of-the-art alternatives, while uniquely satisfying all twelve evaluated security properties.

Incorrect brushing methods normally lead to poor oral hygiene, and result in severe oral diseases and complications. While effective brushing can address this issue, individuals often struggle with incorrect brushing, like aggressive brushing, insufficient brushing, and missing brushing. To break this stalemate, in this paper, we proposed LiT, a toothbrushing monitoring system to assess the brushing status on 16 surfaces using the Bass technique. LiT utilizes commercial LED toothbrushes’ blue LEDs as transmitters, and incorporates only two low-cost photodetectors as receivers on the toothbrush head. It is challenging to determine optimal deployment positions and minimize photodetectors number to establish the light transmission channel in oral cavity. To address these challenges, we established mathematical models within the oral cavity based on the two photodetectors’ deployment to theoretically validate the feasibility and prove robustness. Furthermore, we designed a comprehensive framework to fight against the implementation challenges including brushing action separation, light interference on the outer surfaces of front teeth, toothpaste diversity, user variations, brushing hand variability, and incorrect brushings. Experimental results demonstrate that LiT achieves a highly accurate surface recognition rate of 95.3%, an estimated error for brushing duration of 6.1%, and incorrect brushing detection accuracy of 96.9%. Furthermore, LiT retains stable capability under a variety of circumstances, such as various lighting conditions, user movement, toothpaste diversity, and left and right-handed users.

Owing to the cost of collecting labeled sensor data, self-supervised learning (SSL) methods for human activity recognition (HAR) that effectively use unlabeled data for pretraining have attracted attention. However, applying prior SSL to COMPLEX activities in real industrial settings poses challenges. Despite the consistency of work procedures, varying circumstances, such as different sizes of packages and contents in a packing process, introduce significant variability within the same activity class. In this study, we focus on sensor data corresponding to characteristic and necessary actions (sensor data motifs) in a specific activity such as a stretching packing tape action in an assembling a box activity, and propose to train a neural network in self-supervised learning so that it identifies occurrences of the characteristic actions, i.e., Motif Identification Learning (MoIL). The feature extractor in the network is subsequently employed in the downstream activity recognition task, enabling accurate recognition of activities containing these characteristic actions, even with limited labeled training data. The MoIL approach was evaluated on real-world industrial activity data, encompassing the state-of-the-art SSL tasks with an improvement of up to 23.85% under limited training labels.

Resting-state functional magnetic resonance imaging (rs-fMRI) has significantly advanced the diagnosis of brain diseases. However, existing methods are generally limited to small, disease-specific datasets with less convincing outcomes or lack the interpretability needed to identify reliable disease-associated biomarkers. In this paper, we introduce a novel generative inference model that integrates a Variational Autoencoder (VAE) with Non-negative Matrix Factorization (NMF). Our model comprises three key components: an encoder for learning spatiotemporal dynamic feature embeddings within fMRI data, a decoder to reconstruct the input data from the encoded latent space, and a classifier to distinguish between neurological disorders and normal controls. The three components are simultaneously optimized to perform inference by estimating the posterior distribution of the latent variables from the input fMRI, yielding predictive and interpretable biomarkers for the diagnosis of neurological disorders. We extensively evaluated our method for identifying Autism Spectrum Disorder (ASD) and Alzheimer’s Disease (AD) using two public datasets, ABIDE and ADNI. Experimental results show that our method achieves state-of-the-art performance across various metrics.

The integration of 5G/6G networks with intelligent healthcare systems has enabled early disease detection through patient data monitoring. However, the Internet of Medical Things (IoMT) and remote healthcare services introduce significant privacy and security risks. In this paper, we propose LiteCrypt, which addresses these challenges by introducing an optimized Homomorphic Convolutional Neural Networks (HCNN) structure for secure inference and a lightweight Threshold Signature Scheme (TSS) based dual-authorization mechanism. To enhance the practicality of Homomorphic Encryption (HE)-based secure inference in telemedicine applications, LiteCrypt presents an optimized HCNN framework that ensures efficient and adaptable operations across multiple datasets. A high-performance GPU-accelerated HE engine is developed to address the computational demands of HE operations, enabling real-time processing of encrypted patient data. Besides, LiteCrypt introduces a novel TSS-based dual-authorization protocol, requiring consent from both the patient and the hospital to access patient data, thereby mitigating unauthorized access risks. The system adapts to a flexible 2-out-of-3 authorization scheme for emergencies, ensuring timely data retrieval while maintaining security. To overcome the initial challenge of prolonged computation time due to compute-intensive operations, In LiteCrypt, we utilized the lightweight TSS protocol, based on Oblivious Transfer (OT), which is designed for resource-constrained IoMT devices, reducing computation time from 11.9 to 0.11 seconds. Empirical validation demonstrates LiteCrypt’s superior performance, achieving a 233-fold increase in processing speed, a $96 \%$ reduction in encrypted message size, and a 28-fold speed increase using GPUs.

Keystroke dynamics has proven to be highly effective, with its applications expanding significantly over the years in areas such as preventing transaction fraud, account takeovers, and identity theft. Key-positioning and feature-learning methods are commonly used to identify keystroke signals. However, the existing methods face challenges in detecting overlapping keystrokes and environmentally changed signals. We propose a solution called Chameleon to address these limitations. Unlike previous signal separation and deep learning methods that are ineffective in keystroke signals and computationally demanding, Chameleon employs a low-computation Ranking Model to separate overlapping keystroke signals. Moreover, our experiments demonstrate that Chameleon separated signals can be recognized with an average accuracy of 92.69%, surpassing the commonly used FastICA method, which only reaches 25% accuracy. To account for environmental changes, we utilize the Fréchet Inception Distance (FID) as a guiding metric for model migration. Additionally, we introduce the Inductive Vector, which enables our key-identifying model to adapt to altered environmental conditions such as environment, phone location, and user variety. The Inductive Vector adjusts the model parameters based on the shift in FID. In scenarios with various phone locations, the Inductive Vector significantly improves recognition accuracy from 61% to 98%, outperforming the best existing keystroke recognition algorithm. In other dynamic environmental conditions, our approach achieves an average accuracy rate of 81.7%, which is at least 1.6 times better than the current state-of-the-art keystroke recognition algorithm.

Federated learning (FL) offers distributed machine learning on edge devices. However, the FL model raises privacy concerns. Various techniques, such as homomorphic encryption (HE), differential privacy, and multiparty cooperation, are used to address the privacy issues of the FL model. Among them, HE ensures greater security and privacy since end-to-end encryption maintains data privacy throughout the computation process. Compared with other privacy-preserving techniques, HE does not require the establishment of a trusted environment or protocol among multiple parties and does not involve any artificial noise that can impair system performance. Unfortunately, it suffers from efficiency overhead when applied to privacy-preserving FL (PPFL). Some existing surveys on PPFL discuss the generic construction and organization of PPFL from the perspective of practical HE deployment in PPFL. However, none of them covers the efficiency optimization of HE when applied to PPFL. This article conducts a comprehensive review of the efficiency optimization of HE when applied to PPFL. First, we review general optimization strategies and discuss their limitations when applied directly to HE-based PPFL. Second, an overview of algorithmic, hardware, and hybrid optimizations is provided, along with a discussion of their adaptation. Additionally, we provide a detailed taxonomy of optimizations. Finally, we suggest future HE-based PPFL research directions.

Biochemical information sensing has always been one of the challenges in ubiquitous sensing research for mobile computing. Microorganisms will cause undetectable deterioration in drink production, such as wine and beverage, and microbial contamination is highly susceptible during storage like some liquors can be bottled for sometimes over ten years. Microbial culture methods are common for quality monitoring but unsuitable for real-time beverage quality monitoring. As far as we know, we are the first to use ubiquitous sensing for real-time microbial contamination detection. We designed a lightweight monitoring system called Microbe-Radar, which uses light signals to monitor real-time beverage quality. Microbe-Radar uses eight LEDs and a photodiode to detect fine-grained surface tension and absorption spectrum changes caused by microbial metabolites and growth during deterioration. Characteristic offset degree measurement and absorption spectrum dimension expansion are two critical technologies. Moreover, we implemented countermeasures against ambient light noise and sloshing interference. Microbe-Radar's surface tension and absorption spectrum measurement errors are only 0.89 mN/m and 2.4%, respectively, making identifying the contamination duration, microorganism content, and microorganism composition worthwhile. Experiments showed Microbe-Radar could determine potential issues with liquor quality when the liquid becomes health-threatening or even just contaminated, with an accuracy of 97.5%. Microbe-Radar can also be extended to beverage deterioration warning, with deterioration prediction accuracy of more than 90.6% for five beverages (milk, apple juice, etc.).

The use of speakers in electronic devices has become widespread, but the security risks associated with micro-speakers, such as earphones, are often overlooked. Many assume that soundproof barriers can prevent sound leakage and protect privacy. This article presents the prototype MagEar, an eavesdropping system that exploits magnetic side-channel signals leaked by a micro-speaker to restore intelligible human speech. MagEar outperforms some high-precision magnetometers in detecting magnetic fields at the nanotesla level. Even at a distance of 60 cm, it can recover high-quality audio with a 90% similarity to the original audio. Moreover, the MagEar prototype is portable and can be concealed within a headset housing. We have implemented MagEar as a proof-of-concept system and conducted multiple case studies on the eavesdropping of various speaker-embedded devices, including earphones. The recovered speech can be transcribed using automatic speech recognition techniques, even when obstructed by soundproof walls. It is our aspiration that our work can prompt manufacturers to reconsider the security vulnerabilities of speakers.

Neglecting proper oral hygiene has proven to potentially lead to severe oral disease, resulting in complications over time. Careful brushing can mitigate the problem, but it is common for individuals to dedicate insufficient time to the various areas of their teeth. We propose LiT to monitor the brushing situation of 16 Bass technique surfaces in real-time. LiT relies on commercial toothbrushes with blue LEDs as a transmitter and requires only 2 low-cost photosensors as receivers on the toothbrush head. However, the transmission channel of light in the oral cavity is unclear. Finding the optimal deployment positions and minimizing the number of photosensors is challenging. To tackle these obstacles, we design the positioning of the 2 photosensors and create a transmission model within the oral cavity to verify the feasibility theoretically. Additionally, obstacles in implementation include separating brushing action accurately, interference of light on the outer surfaces of front teeth, and individual variability. To overcome these challenges, we develop corresponding technologies and a comprehensive framework. Experiments with 16 users show that LiT achieves a highly accurate recognition rate of 95.3% with an error estimate for brushing duration of 6.1%. Furthermore, LiT also proves resilient under user motion and environmental interference.

Traditional liquid identification instruments are often unavailable to the general public. This paper shows the feasibility of identifying unknown liquids with commercial lightweight devices, such as a smartphone. The wisdom arises from the fact that different liquid molecules have various viscosity coefficients, so they need to overcome dissimilitude energy barriers during relative motion. With this intuition in mind, we introduce a novel model that measures liquids’ viscosity based on active vibration. The idea sounds straightforward, yet, it is challenging to build up a robust system utilizing the built-in accelerometer in smartphones. Practical issues include under-sampling, self-interference, and volume change impact. Instead of using machine learning techniques, we tackle these issues through multiple signal processing stages to reconstruct the original signals and cancel out the interference. Our approach achieved the liquid viscosity estimates with a mean relative error of 2.3% and distinguish 30 kinds of liquid with an average accuracy of 97.33%.

Speakers have been widely embedded in various electronic devices as a standard configuration. The security vulnerability of microspeakers (such as earphones) is commonly overlooked because it is often assumed that soundproof boundaries, such as walls, can prevent privacy-infringing sound leakage. In this paper, we present the prototype MagEar, an eavesdropping system that leverages magnetic side-channel signals leaked by a microspeaker to recover intelligible human speech. MagEar has sufficiently high sensitivity to detect magnetic fields on the order of nanotesla, exceeding some high-precision magnetometers. It can recover high-quality audio with 90% similarity to the original audio even at a distance of 60 cm. In addition, the MagEar prototype is portable and can be hidden in a headset shell. We have implemented MagEar as a proof-of-concept system and conducted several case studies of eavesdropping on different types of speaker-embedded devices, including earphones, and we have demonstrated the ability to successfully transcribe the recovered speech using automatic speech recognition techniques even when blocked by soundproof walls. We hope that our work can push manufacturers to rethink this security vulnerability of speakers.

In industrialized wine production, brewing and aging are two key steps. These two processes require the liquors to be bottled for a long time, sometimes more than ten years. The liquor is vulnerable and highly susceptible to microbial contamination during storage, causing undetectable deterioration. During the production process, wineries control the indoor temperature and carbon dioxide concentration to slow down other microorganisms' reproduction speed. These methods, however, do not prevent pathogenic microorganism growth. Currently, microbial culture methods are not suitable for real-time liquor quality monitoring in wineries. Therefore, we have designed a lightweight monitoring system called Lili, which uses light signals to monitor real-time liquor quality changes. Lili detects the changes in surface tension and absorption spectrum caused by microbial metabolites and growth during deterioration. Lili employs eight LEDs and one photodiode to achieve fine-grained surface tension and absorption spectrum measurements. By analyzing these changes, Lili realizes real-time quality monitoring. In this paper, the characteristic offset degree measurement and the absorption spectrum dimension expansion are two critical technologies. In addition, we implemented countermeasures against ambient light noise and sloshing interference. Lili's surface tension and absorption spectrum measurement errors are only 0.89 mN/m and 2.4%, respectively, making it useful to identify the contamination duration, microorganism content and microorganism composition. These two data points can be used to determine potential issues with liquor quality when the liquor becomes health-threatening or even just contaminated, with an accuracy of 97.5%.

Traditional liquid identification instruments are often unavailable to the general public. This paper shows the feasibility of identifying unknown liquids with commercial lightweight devices, such as a smartphone. The wisdom arises from the fact that different liquid molecules have various viscosity coefficients, so they need to overcome dissimilitude energy barriers during relative motion. With this intuition in mind, we introduce a novel model that measures liquids' viscosity based on active vibration. Yet, it is challenging to build up a robust system utilizing the built-in accelerometer in smartphones. Practical issues include under-sampling, self-interference, and volume change impact. Instead of machine learning, we tackle these issues through multiple signal processing stages to reconstruct the original signals and cancel out the interference. Our approach could achieve the liquid viscosity estimates with a mean relative error of 2.9% and distinguish 30 kinds of liquid with an average accuracy of 95.47%.

Atmospheric humidity is a significantly important factor in our daily lives, as it is closely bound up with agriculture, industrial production, human health and so on. Therefore, efficient and precise humidity measurement techniques are indispensable. However, the existing off-the-shelf techniques, including the dry and wet bulb hygrometer, humidity sensor as well as WiFi based detector, all fail to achieve a sensitive, accurate and convenient humidity measurement, especial for a large scale deployment. In this paper, we observe that different levels of water vapor have certain impact on millimeter wave (mmWave) signals in indoor environments. Accordingly, we propose an fine-grained environmental humidity sensing technology via wireless signals in the mmWave band. However, mmWave signals are not only sensitive to humidity, but also other environmental factors, such as oxygen. To establish a linear relationship between humidity and mm Wave signal propagation, we exploit a subspace projection technique to remove the environmental noise. Upon extracting the humidity-associated features in the noise-free signal, we utilize support vector machine (SVM) to model the humidity measurement classifier of a certain place. Extensive experiments have been conducted in different scenarios in order to verify the effectiveness of the proposed system. Results show that the average accuracy of humidity measurement is up to 85 % when the humidity interval is 3 %, and is 95 % when the humidity interval is 5%. We further show that the proposed method is very sensitive to the humidity dynamics and is 63.2 times faster compared to the traditional hygrometers.

The emergence of wearable devices has brought great simplicity and convenience to people's daily lives. However, due to the small form-factor, low-profile hardware interfaces, the input scheme for such wearable devices becomes a bottleneck and even sabotages their functionalities. The state-of-the-art interaction schemes, including voice input, inertial measurement unit (IMU)based input, or acoustic based input, all require a stable environment, which is critical for wearable device. To break this stalemate, we propose a stable QWERTY keyboard input for wearable devices based on bone-conduction models. Using the characteristics of human anatomy, we achieve a low-cost and high precision text input system, named Osteoacusis input (Oinput), with the help of human bones. To be specific, we first investigate a new set of bone-conduction theories. Through this set of theories, we combine a strong anti-noise cyclic neural network to achieve a high-precision QWERTY keyboard recognition for text input. Furthermore, in order to improve the user experience, we leverage slightly keyboard layout changing, dimensionality and feature selection to reduce the power consumption while preserving the convenience and stability. We have conducted experiments on 30 volunteers. The results show that Oinput has superior robustness with a high recognition accuracy of 93.3% in average. Moreover, Oinput's calibration mechanism increases the accuracy by more than 99%.

Conflagration is one of the major disasters that threatens human life and property. If the proper action is not taken in detecting the symptom of conflagration events ahead of time, the number of such disasters will keep increasing. An effective solution in this context will alleviate many fire-related global problems to a great extent. Although fire detectors are not available in many places, WiFi networks are increasingly prevalent nowadays. Motivated by the previous works that used WiFi signals for the purpose of environment monitoring and activity recognition, we make an attempt to use WiFi signals to detect fire. Through several experiments, we find that fire influences the transmission of wireless signals uniquely, and consequently it affects the amplitude and phase of the resultant Channel State Information (CSI). Based on this observation, in this paper, we propose a device-free fire detection system, namely Wi-Fire, using commercial WiFi devices. To the best of our knowledge, this is the first work that leverages CSI of radio frequency (RF) signal to detect fire events using existing wireless infrastructure without requiring any additional device. We implement our proposed system on desktop computers equipped with commercial 802.11n network interface cards (NICs). Comprehensive experiments have been conducted for different scenarios in different environments to verify the effectiveness of our proposed system. The results verify that the fire detection accuracy of this training-based system is up to 96.67% on average.