Deep learning for cancer intelligent diagnosis based on multi-omics data has achieved enormous advance in the condition of sufficient samples. However, these methods can’t generalize to circumstances that certain cancer samples are few, which poses a challenge for few-shot learning to apply in it. Therefore, we propose a few-shot learning algorithm, called the multi-omics feature reconstruction (MOFR), to realize the cancer classification and the metastasis prediction. Specifically, MOFR transforms the classification problem into a feature reconstruction issue in latent space. This approach needs fewer learning parameters, which facilitates model fitting in such environment. By directly regressing the ’distance’ between the support feature set and the query feature, it achieves classification tasks. Additionally, we also design a tailored feature extractor to effectively map raw data into high-level and informative features. It draws inspiration from the Transformer architecture but overcomes the need for extensive training data to train the model. According to extensive experimental results, the proposed few-shot learning algorithm surpasses homogeneous approaches in both classification and metastasis prediction performance. We provide detailed code for the new model proposed in this paper on https://github.com/vbfeobveofn/Code.git .

The open nature of wireless communications renders unmanned aerial vehicle (UAV) communications vulnerable to impersonation attacks, under which malicious UAVs can impersonate authorized ones with stolen digital certificates. Traditional fingerprint-based UAV authentication approaches rely on a single modality of sensory data gathered from a single layer of the network model, resulting in unreliable authentication experiences, particularly when UAVs are mobile and in an openworld environment. To transcend these limitations, this paper proposes SecureLink, a UAV authentication system that is among the first to employ cross-layer information for enhancing the efficiency and reliability of UAV authentication. Instead of using single modalities, SecureLink fuses physical-layer radio frequency (RF) fingerprints and application-layer micro-electromechanical system (MEMS) fingerprints into reliable UAV identifiers via multimodal fusion. SecureLink first aligns fingerprints from channel state information measurements and telemetry data, such as feedback readings of onboard accelerometers, gyroscopes, and barometers. Then, an attention-based neural network is devised for in-depth feature fusion. Next, the fused features are trained by a multi-similarity loss and fed into a one-class support vector machine for open-world authentication. We extensively implement our SecureLink using three different types of UAVs and evaluate it in different environments. With only six additional data frames, SecureLink achieves a closed-world accuracy of 98.61% and an open-world accuracy of 97.54% with two impersonating UAVs, outperforming the existing approaches in authentication robustness and communication overheads. Finally, our datasets collected from these experiments are available on GitHub: https://github.com/PhyGroup/SecureLink_data.

Federated learning (FL) enables collaborative intrusion detection in the Internet of Vehicles (IoV) without centralizing sensitive vehicular data, yet current FL-based intrusion detection systems (IDS) remain limited: they lack formal privacy guarantees against gradient inversion attacks, treat all participating vehicles uniformly despite heterogeneous threat exposure, and incur high communication overhead from transmitting full model gradients over bandwidth-constrained V2X links. We present FedSADP-IDS, a framework that addresses these gaps through three coordinated mechanisms. First, a self-adaptive differential privacy module adjusts the per-round privacy budget ε according to local threat levels—injecting less noise when a vehicle is under active attack so that attack-discriminative gradients are preserved, and tightening protection during normal operations. Second, a Fisher Information Matrix-guided gradient selection scheme, weighted by attack-type risk scores, uploads only the top-K% most security-relevant parameters, cutting communication by 40% with no accuracy penalty. Third, FedRisk, a risk-weighted aggregation strategy, up-weights contributions from vehicles experiencing high-severity attacks to improve the global model’s sensitivity to rare but critical threat patterns. The training objective also incorporates a risk-weighted supervised contrastive loss (RW-SupCon) that encourages tighter feature clusters for high-risk classes and wider margins between classes of disparate severity. On the device side, LightGuard—alightweight extractor built on depthwise separable convolutions with multi-scale temporal branches and channel attention—runs at 4.2 ms per inference on a Raspberry Pi 4. Evaluated on the Car-Hacking, CAN-Intrusion, and CICIDS2017 datasets, FedSADP-IDS reaches 97.6% accuracy under formal per-client (ε, δ)-differential privacy (ε ≈ 3.8 in the worst case across all clients), improving over existing federated IDS methods by 0.4–3.4% in accuracy while reducing communication cost.

Due to the openness of the wireless medium, smartphone users are susceptible to user privacy attacks, where user privacy information is inferred from encrypted Wi-Fi wireless traffic. Existing attacks are limited to recognizing mobile apps and their actions and cannot infer the smartphone user identity, a fundamental part of user privacy. To overcome this limitation, we propose U-Print, a novel attack system that can passively recognize smartphone apps, actions, and users from over-the-air MAC-layer frames. We observe that smartphone users usually prefer different add-on apps and in-app actions, yielding different changing patterns in Wi-Fi traffic. U-Print first extracts multi-level traffic features and exploits customized temporal convolutional networks to recognize smartphone apps and actions, thus producing users' behavior sequences. Then, it leverages the silhouette coefficient method to determine the number of users and applies the k-means clustering to profile and identify smartphone users. We implement U-Print using a laptop with a Kali dual-band wireless network card and evaluate it in three real-world environments. U-Print achieves an overall accuracy of 98.4% and an F1 score of 0.983 for user inference. Moreover, it can correctly recognize up to 96% of apps and actions in the closed world and more than 86% in the open world.

Lung cancer is a potentially fatal disease worldwide, and improving the accuracy of diagnosis plays a key role in enhancing patient outcomes. In this study, we extended computer-aided work to the task of assisting tracheoscopy in predicting lung cancer subtypes. To solve the problem of information fusion in different spatial scales and channels, we proposed MrgaNet. The network enhances classification performance by expanding interactions from low to high orders, dynamically adjusting feature weights, and incorporating a channel competition operator for efficient feature selection. Our network achieved a precision of 0.87 in the endobronchial dataset. In addition, the accuracy of 89.25% and 96.76% was achieved in the Kvasir-v2 dataset and the Kvasir-Capsule dataset, respectively. The results demonstrate that MrgaNet achieves superior performance compared to existing excellent methods.

With the increasing emphasis on the security of satellite internet communications, growing attention has been directed toward inter-satellite link (ISL)-based routing issues. Due to the instability and openness of inter-satellite links, Low Earth Orbit (LEO) satellite networks are highly vulnerable to malicious attacks, which can result in significant degradation of network performance. To address this problem, this paper proposes a novel trust-aware and resilience-oriented secure routing (TARSR) algorithm to improve the robustness of path selection. The algorithm incorporates a robust framework that integrates node trust evaluation, a multipath computation mechanism, and a link security assessment model. Finally, the simulation results verified that compared to the existing routing protocols, the proposed algorithm results in a lower packet loss rate and reduced end-to-end latency.

Over-the-air computation (OAC) can leverage the interference between multiple users of multi-access channels (MAC) to perform function computation during the transmission process. This helps overcome the high latency issues caused by a large amount of data access in the applications of Internet of Things (IoT). The majority of current studies solely focus on air computation with the assumption of full channel state information (CSI). However, in practical applications, due to quantization and feedback errors, calibration mismatches, and various factors leading to delay errors and noise, the central end often can only obtain imperfect CSI. Therefore, we have a strong interest in optimizing the design of OAC with channel estimation errors to enhance its performance under adverse channel conditions. This papeer suggests a superior design strategy for OAC, considering total power limitations and flawed CSI. Specifically, we formulate the problem as a minimum mean square error (MSE) and transform the original non-convex problem into a convex problem using alternating optimization methods, and obtain the closed-form solution for the optimal transceiver factors. Numerical results are presented to validate the effectiveness of the proposed method.

The security risks posed by illegal drones impersonating legal drones to carry out illegal activities is increasing. Drone identification technology can be used to identify these masquerading drones to avoid accidents. Traditional radio frequency (RF) and radar drone identification technologies suffer from limitations such as susceptibility to interference, high cost of additional equipment, and difficulty in deployment. Drone identification using optics or acoustics is heavily influenced by environmental conditions. To address these challenges, this paper proposes a novel drone identification system. It takes micro-electromechanical system (MEMS) sensor readings obtained from drone telemetry data and then utilizes a convolutional neural network (CNN) to extract features as drone fingerprint for identification. This method is environment-independent and requires no additional equipment. We implement the system and evaluate it with ten DJI Tello drones, the experimental results show that the system using 36 sensor data frames is able to achieve more than 97% identification accuracy.

The rapid proliferation of Internet of Things (IoT) devices has significantly expanded the network attack surface, necessitating the deployment of advanced AI (artificial intelligence)-based intrusion detection systems (IDS) to bolster IoT security. But existing methods face two significant challenges: (1) Feature redundancy: Current approaches extract numerous flow-level features to learn attack behavior, resulting in high computational complexity and substantial redundant information. (2) Class imbalance: Limited attack traffic samples hinder models from effectively learning attack patterns. However, existing algorithms typically address only one of these issues, overlooking their interconnection. Therefore, we propose a Feature Selection and Large Language Models (LLMs)-based IoT intrusion detection framework (FSLLM). At its core is a multi-stage feature selection algorithm combining Minimum Redundancy Maximum Relevance algorithm (mRMR) and a Pearson Correlation Coefficient (PCC)-improved Covariance Matrix Adaptation Evolution Strategy algorithm (CMA-ES). This algorithm utilizes the CMA-ES algorithm for feature search while also taking into account the mutual information and collinearity among features, thereby more effectively reducing redundancy features. Subsequently, we employ the selected representative features to fine-tune LLMs and generate additional attack samples. This approach effectively reduces the computational cost of fine-tuning while producing higher-quality samples. Furthermore, we employ Focal Loss (FL) function-improved LightGBM as the classifier to improve detection performance. We evaluate our framework on five IoT intrusion detection datasets: NF-ToN-IoT-v2, NF-UNSW-NB15-v2, NF-BoT-IoT-v2, NF-CSE-CIC-IDS2018-v2, and CIC-ToN-IoT. Experimental results demonstrate that FSLLM achieves comparable or superior accuracy to current state-of-the-art methods while reducing redundant features by over 80%.

For immersive 3D reconstruction due to its omnidirectional scene understanding capacity. Prevailing methods primarily employ projection fusion paradigms between equirectangular projection (ERP) and cubic maps, we still need to confront two inherent limitations: 1) progressive distortion accumulation during spherical-to-polyhedral conversion, particularly severe at high latitude regions, and 2) feature discontinuity artifacts induced by incompatible sampling densities across projection boundaries. We present GFusion, an efficient framework that synergizes ERP images with Icosahedral Projection (ICO) through three innovative components: (1) A geometry-preserving converter establishing continuous ERP-to-ICOSAP mapping via adaptive spherical tessellation; (2) GDC-ResNet architecture combining gated depth-wise convolution with boundary-sensitive attention to resolve multi-scale discontinuities; (3) A Global Geometric Awareness Fusion (GAF) module enabling bidirectional feature interaction between ERP local details and ICOSAP global structures. Comprehensive evaluations demonstrate that our approach achieves competitive performance against leading contemporary methods while maintaining superior computational efficiency.

Cyber resilience has attracted worldwide attention. Security and reliability of information transmission are considered to be two key characteristics from the perspective of physical layer cyber resilience. Consequently, a dynamic secure polar code is proposed in this paper to improve the physical layer resilience of the non-degraded channels, in which a shared key is firstly used as a seed to produce a pseudo-random sequence that is further used to generate a dynamic generator matrix to obfuscate the coding structure. Then, a small amount of random column vectors are inserted into the generator matrix to destroy the systematicness of the code. Finally, the hash value of the key is used as frozen bits to avoid the leaking of the key. Confidentiality analysis and computer simulations were carried out for the proposed scheme. The results show that the proposed scheme has high security against cryptanalysis attacks. Meanwhile the eavesdropper’s bit error rate is kept above 0.435 in non-degraded channel scenarios, and the reliability is almost the same with that of the original polar code. Furthermore, the frame error rate of the proposed scheme is about 1/10 of the existing schemes in low signal-to-noise ratio regimes.

Image stitching is a crucial task in computer vision, enabling the creation of panoramic images for augmented reality/virtual reality (AR/VR) experiences. Image registration plays an important role in achieving spatial continuity and consistency in image stitching. Therefore, achieving accurate image registration is critical to high-quality panoramic image generation. Although recent learning-based methods have improved the optimization of stitched images, they still rely on simple associations of images or feature maps to search for feature correspondences, neglecting the capture of these correspondences during feature extraction. Additionally, directly warping the original image with the homography matrix extracted from the feature space ignores the network's influence on image transformation relationships. To address these limitations and enhance the quality of panoramic images, we propose an artificial intelligence (AI) image registration network based on a cross-attention mechanism. Our approach incorporates a Local Transformer to help the network perceive image correspondences. Furthermore, we impose constraints on network training through warp-equivariance to mitigate the impact on image transformation relationships. These strategies increase the accuracy and generalization of the AI-based registration method and form a better image stitching algorithm. Finally, we apply the AI image stitching algorithm to construct a panoramic stereoscopic live broadcast system. Experimental results show that our method achieves competitive results and satisfies the requirements of panoramic systems.

The rapid advancement of technologies such as virtual and augmented reality has garnered substantial attention from both academia and industry. As the foundation of immersive multimedia content, omnidirectional image generation necessitates the development of robust and efficient image stitching algorithms. Unlike planar image stitching, omnidirectional image stitching involves stitching images captured by binocular or quadrinocular camera systems, introducing lager parallaxes (e.g., 90 degrees or even 180 degrees), which lead to pronounced distortions and hardto-remove artifacts. Conventional planar image stitching methods based on optical flow typically employ unidirectional models. However, the intrinsic spatial relationships between sub-images in omnidirectional image necessitate the use of bidirectional optical flow. Existing approaches often estimate optical flow for each branch independently through simple feature mapping, neglecting the latent correlations between bidirectional flow. To address these challenges, we propose using a seam-driven approach to replace the traditional weighted blending strategy to effectively minimize artifacts. Additionally, we incorporate an advanced attention mechanism to establish a "bridge" that enables joint optimization of the two optical flow branches. Finally, we lightweight the model at a small cost of precision. Experimental results demonstrate that our method comprehensively outperforms the baseline model and generates natural omnidirectional images.

Radio frequency (RF) fingerprinting techniques provide a promising supplement to cryptography-based approaches but rely on dedicated equipment to capture in-phase and quadrature (IQ) samples, hindering their wide adoption. Recent advances advocate easily obtainable channel state information (CSI) by commercial WiFi devices for lightweight RF fingerprinting, while falling short in addressing the challenges of coarse granularity of CSI measurements in an open-world setting. In this paper, we propose CSI2Q, a novel CSI fingerprinting system that achieves comparable performance to IQ-based approaches. Instead of extracting fingerprints directly from raw CSI measurements, CSI2Q first transforms frequency-domain CSI measurements into time-domain signals that share the same feature space with IQ samples. Then, we employ a deep auxiliary learning strategy to transfer useful knowledge from an IQ fingerprinting model to the CSI counterpart. Finally, the trained CSI model is combined with an OpenMax function to estimate the likelihood of unknown ones. We evaluate CSI2Q on one synthetic CSI dataset involving 85 devices and two real CSI datasets, including 10 and 25 WiFi routers, respectively. Our system achieves accuracy increases of at least 16% on the synthetic CSI dataset, 20% on the in-lab CSI dataset, and 17% on the in-the-wild CSI dataset.

The proliferation of time-sensitive Internet of Things (IoT) applications has significantly increased the demand for real-time communication in uplink orthogonal frequency division multiple access (UL-OFDMA) Wi-Fi networks. Despite extensive studies, how to meet the heterogeneous Age of Information (AoI) requirements across different stations (STAs) in Wi-Fi networks remains an open question. To tackle this issue, we propose a multiagent reinforcement learning (MARL) resource allocation algorithm based on independent hybrid proximal policy optimization (IHPPO), aiming to minimize the AoI and power consumption for each STA while guaranteeing the heterogeneous AoI requirements among STAs within a resource-constrained environment. Specifically, the proposed strategy utilizes hybrid PPO (HPPO) to directly optimize the original hybrid action space by combining discrete resource unit (RU) selection and continuous transmit power adjustment. Extensive simulations demonstrate the superior performance of the IHPPO mechanism in terms of convergence performance and the tradeoff between AoI and power consumption, relative to the decomposed multiagent deep deterministic policy gradient (DE-MADDPG) algorithm, the fully decentralized MADDPG (FD-MADDPG) algorithm, and the random method in different access scenarios.

The growing complexity of cybersecurity systems and the escalating severity of security threats have rendered the concept of traditional absolute security impractical. As an alternative paradigm, cyber resilience has emerged as a critical issue which focus on ensuring business continuity and system availability. At the meantime, researches on quantitative evaluation of cyber resilience also attract worldwide attention. This paper proposes a quantitative multi-metric fusion method for comprehensive cyber resilience evaluation, which integrates multiple key performance metrics to provide a comprehensive evaluation. Specifically, multiple metrics are measured separately under various perturbations, and then the entropy weight method is used to determine the weight of each metric. Additionally, a revised 2-additive fuzzy measures method is proposed to figure out the interactions among different metrics. The resilience value of each metric is quantified by the area under the curve (AUC) method. Finally, fuzzy integration is applied to synthesize these values into a comprehensive cyber resilience score. The proposed method is executed on a Web server system to analyze its comprehensive cyber resilience across distinct perturbation scenarios. Experimental results demonstrate that the proposed method not only accurately captures the comprehensive cyber resilience across several individual dimensions, but also effectively reveals the complex effects of multi-dimensional interactions on the comprehensive cyber resilience.

Reflecting Intelligent Surface (RIS)-assisted Integrated Sensing and Communication (ISAC) systems can achieve synergistic enhancement of communication and sensing under limited spectrum conditions, where the joint active-passive beamforming design holds significant importance in complex electromagnetic environments. To address the issue of strong clutter interference in scenarios where multi-user communication coexists with target sensing, this paper proposes an efficient joint beamforming optimization algorithm that achieves coordinated optimization of base station beamforming, receiver equalizer, and RIS phase shifts, thereby effectively suppressing clutter interference. Simulation results show that the proposed scheme significantly outperforms benchmark methods in both communication rate and sensing performance.

Traffic prediction is an important component of Intelligent Transportation System (ITS), and accurately predicting dynamic traffic conditions is a challenge to be solved. Capturing the locality and globality of spatiotemporal correlations plays a significant role in prediction, and existing methods cannot completely capture them. This paper proposes a Long Short Term Spatio-temporal Neural Network (LS2TNN) model for traffic speed prediction. The spatial gated graph convolutional attention module and time convolutional attention module in the model are used to capture long-term and short-term spatiotemporal correlations. Static graphs cannot effectively describe dynamic traffic flow. This paper uses predefined edge dynamic edge weights graphs and dynamic edge dynamic edge weights graphs for Graph Convolution Network (GCN), in order to better describe the dynamic traffic situation in real scenes. Experiments conducted on two real datasets showed that LS2TNN achieved advanced performance compared to the baselines.

Identity authentication is crucial for ensuring the information security of wireless communication. Radio frequency (RF) fingerprinting techniques provide a prom-ising supplement to cryptography-based authentication approaches but rely on dedicated equipment to capture in-phase and quadrature (IQ) samples, hindering their wide adoption. Recent advances advocate easily obtainable channel state in-formation (CSI) by commercial WiFi devices for lightweight RF fingerprinting, but they mainly focus on eliminating channel interference and cannot address the challenges of coarse granularity and information loss of CSI measurements. To overcome these challenges, we propose CSI2Q, a novel CSI fingerprinting sys-tem that achieves comparable performance to IQ-based approaches. Instead of ex-tracting fingerprints directly from raw CSI measurements, CSI2Q first transforms them into time-domain signals that share the same feature space with IQ samples. Then, the distinct advantages of an IQ fingerprinting model in feature extraction are transferred to its CSI counterpart via an auxiliary training strategy. Finally, the trained CSI fingerprinting model is used to decide which device the sample under test comes from. We evaluate CSI2Q on both synthetic and real CSI datasets. On the synthetic dataset, our system can improve the recognition accuracy from 76% to 91%. On the real dataset, CSI2Q boosts the accuracy from 67% to 82%.