Chinese textual affective structure analysis (CTASA) is a sequence labeling task that often relies on supervised deep learning methods. However, acquiring a large annotated dataset for training can be costly and time-consuming. Active learning offers a solution by selecting the most valuable samples to reduce labeling costs. Previous approaches focused on uncertainty or diversity but faced challenges such as biased models or selecting insignificant samples. To address these issues, multilevel active learning (MAL) is introduced, which leverages deep textual information at both the sentence and word levels, taking into account the complex structure of the Chinese language. By integrating the sentence-level features extracted from bidirectional encoder representations from Transformers (BERT) embeddings and the word-level probability distributions obtained through a conditional random field (CRF) model, MAL comprehensively captures the Chinese textual affective structure (CTAS). Experimental results demonstrate that MAL significantly reduces annotation costs by approximately 70% and achieves more consistent performance compared to baseline methods.
In recent research on image encryption, many schemes associate the key generation mechanism with the plaintext to resist chosen plaintext attacks. However, when the sender encrypts many images, a large amount of additional data related to the plaintext need to be transmitted, which leads to problems such as high transmission costs, high requirements for key storage space, and complex key management. Therefore, in this paper, we propose a self-sufficient plaintext-related JPEG image encryption scheme based on a unified key (SPJEU). This scheme establishes the connection between the plaintext and the key by selecting the direct current (DC) coefficients in the JPEG image through a unified key. Homomorphic encryption is applied to the selected DC coefficients, allowing plaintext information to be decrypted directly from the ciphertext domain using a specific calculation method. The remaining DC coefficients are encrypted through group diffusion, and the alternating current (AC) coefficients are grouped and permuted based on the run length. Extensive experiments show that our scheme can resist chosen plaintext attacks, avoid transmitting plaintext-related additional data in the communication channel, and simplify key management. This scheme also ensures the security and format compatibility of the ciphertext image, and the file increment after encryption is very small.
Deep learning-based intrusion detection systems rely on numerous training samples to achieve satisfactory detection rates. However, in the real-world Internet of Things (IoT) environments, the diversity of IoT devices and the subsequent fragmentation of attack types result in a limited number of training samples, which urgently requires researchers to develop few-shot intrusion detection systems. In this study, we propose a subspace-based approach for few-shot IoT intrusion detection systems to cope with the dilemma of insufficient learnable samples. The method is based on the principle of classifying metrics to identify network traffic. After feature extraction of samples, a subspace is constructed for each category. Next, the distance between the query samples and the subspace is calculated by the metric module, thus detecting malicious samples. Subsequently, based on the CICIoT2023 dataset we construct a few-shot IoT intrusion detection dataset and evaluate the proposed method. For the detection of unknown categories, the detection accuracy is 93.52% in the 5-way 1-shot setting, 92.99% in the 5-way 5-shot setting, and 93.65% in the 5-way 10-shot setting.
This study presents a parallel version of the string matching algorithms research tool (SMART) library, implemented on NVIDIA's compute unified device architecture (CUDA) platform, and uses general-purpose computing on graphics processing unit (GPGPU) programming concepts to enhance performance and gain insight into the parallel versions of these algorithms. We have developed the CUDA-enhanced SMART (CUSMART) library, which incorporates parallelized iterations of 64 string matching algorithms, leveraging the CUDA application programming interface. The performance of these algorithms has been assessed across various scenarios to ensure a comprehensive and impartial comparison, allowing for the identification of their strengths and weaknesses in specific application contexts. We have explored and established optimization techniques to gauge their influence on the performance of these algorithms. The results of this study highlight the potential of GPGPU computing in string matching applications through the scalability of algorithms, suggesting significant performance improvements. Furthermore, we have identified the best and worst performing algorithms in various scenarios.
Recent attention to privacy issues demands a communication-safe method for training human activity recognition (HAR) models on client activity data. Federated learning (FL) has become a compelling technique to facilitate model training between the server and clients while preserving data privacy. However, classical FL methods often assume independent and identically distributed (IID) data among clients. This assumption does not hold true in practical scenarios. Human activity in real-world scenarios varies, resulting in skewness where identical activities are executed uniquely across clients. This leads to local model objectives drifting away from the global model objective, thereby impeding overall convergence. To address this challenge, we propose FedCoad, a novel federated model leveraging contrastive learning with adaptive control variates to handle the skewness among HAR clients. Model contrastive learning minimizes the gap in representation between global and local models to help global model convergence. During local model updates, the adaptive control variates penalize the local model updates with respect to the model weight and the rate of change from the control variates update. Our experiments show that FedCoad outperforms state-of-the-art FL algorithms on HAR benchmark datasets.
Recently, large-scale pretrained models have revealed their benefits in various tasks. However, due to the enormous computation complexity and storage demands, it is challenging to apply large-scale models to real scenarios. Existing knowledge distillation methods require mainly the teacher model and the student model to share the same label space, which restricts their application in real scenarios. To alleviate the constraint of different label spaces, we propose a prototype-guided cross-task knowledge distillation (ProC-KD) method to migrate the intrinsic local-level object knowledge of the teacher network to various task scenarios. First, to better learn the generalized knowledge in cross-task scenarios, we present a prototype learning module to learn the invariant intrinsic local representation of objects from the teacher network. Second, for diverse downstream tasks, a task-adaptive feature augmentation module is proposed to enhance the student network features with the learned generalization prototype representations and guide the learning of the student network to improve its generalization ability. Experimental results on various visual tasks demonstrate the effectiveness of our approach for cross-task knowledge distillation scenarios.
Image steganography algorithms based on deep learning are often trained using either spatial- or frequency-domain features. It is difficult for features from a single domain to comprehensively express the content of an entire image, which usually leads to poor performance because steganography is commonly multi-task. To solve this problem, this paper proposes a robust image steganography algorithm based on feature score maps, called the secure and robust image steganography network (SRIS-Net). First, instead of spatial-domain steganography, our proposed algorithm utilizes a convolutional neural network to obtain shallow spatial-domain features. These features are decomposed by Laplacian pyramid frequency-domain decomposition (LPFDD) to hide secret information in the different frequency sub-bands with a progressive assisted hiding strategy that significantly reduces the influence of the secret information on the cover image, achieving significant invisibility and robust performance. In addition, we propose a global-local embedding module (GLEM) to achieve embedding by considering the overall structure of the image and the local details, and a dual multi-scale aggregation sub-network (DMSubNet) to perform multi-scale reconstruction to improve the quality of the carrier image. For security, we propose a dual-task discriminator structure, while giving a real/fake judgment of the image, which can generate a feature score map of the cover image's region of interest (ROI) to guide the embedding module to generate a carrier image with higher imperceptibility and undetectability. Experimental results on BOSSBase show that our SRIS-Net outperforms mainstream methods in terms of undetectability and robustness, with more than 9.2 and 3.4 dB improvement in visual quality, respectively, and the capacity can be increased up to approximately 72-96 bits per pixel.
In recent years, the exploration and application of resistance networks have expanded significantly, and solving the equivalent resistance between two points of a resistance network has been an important topic. In this paper, we focus on optimizing the formula for calculating the two-point resistance of an m × n cobweb resistance network with 2r boundary conditions. To improve the computational efficiency of the equivalent resistance between two points, the formula is optimized by using the optimal approximation property of Chebyshev polynomials in combination with hyperbolic functions, and the derivation process is simplified. We discuss the equivalent resistance formulas in several special cases and compare the computational efficiency of the equivalent resistance formulas before and after optimization. Finally, we make an innovative attempt at path planning through potential formulas and propose a heuristic algorithm based on cobweb potential function for robot path planning in a cobweb environment with obstacles.
High-precision indoor positioning offers valuable information support for various services such as patient monitoring, equipment scheduling management, and laboratory safety. A traditional indoor positioning technology, fingerprint indoor positioning, often employs the K-nearest neighbor (KNN) algorithm to identify the closest K reference points (RPs) via the received signal strength (RSS) for location prediction. However, RSS is susceptible to environmental interference, leading to the selection of RPs that are not physically the closest to the user. Moreover, using a fixed K value is not the optimal strategy. In this work, we propose a novel approach, the dynamic K-nearest neighbor method based on strong access point (AP) credibility (SAPC-DKNN), for indoor positioning. In SAPC-DKNN, we leverage prior knowledge of RSS path loss and employ the RSS fluctuation area to quantify the significance of different APs. We integrate the similarity of AP sets within the range of strong APs and formulate a weighted distance metric for RSS based on the credibility of strong APs. Additionally, we introduce a dynamic K-value algorithm based on neighbor density (ND-DKA) for the automatic optimization of the K value for each test point. Experimental evaluations conducted on three datasets demonstrate that our method significantly reduces the average positioning error by 15.41%-64.74% compared to the state-of-the-art KNN methods.
For the problem of tracking maritime dim targets, the sequential Monte-Carlo multi-Bernoulli track-before-detect (SMC-MB-TBD) method is popular. However, this method may face low tracking accuracy and tracking loss due to particle impoverishment and velocity uncertainty. In this study, a novel filter called position scaling and velocity correction multi-Bernoulli (PSVC-MB) is proposed to deal with this problem. First, particle position scaling is used to replace resampling in the SMC-MB-TBD method to deal with the lack of particle diversity. Second, when the target is stably tracked, the target velocity is extracted from the multi-frame information and used for re-estimation. Pseudo point measurements are calculated from the weighted average of all locations near the particle position, and the particle velocity will be continuously corrected with the pseudo point measurements. Simulation results verify the effectiveness of the proposed method at different low signal-to-clutter ratios (SCRs).
To meet the access demands of massive terminal users, the space-based Internet of Things (IoT) requires sufficient frequency resources for allocation. However, the frequency resources that are currently available have already been allocated to a great extent. Furthermore, the utilization rate of the allocated frequency resources is low. To support massive user access under restricted frequency resources, this work proposes a scheme based on Doppler frequency offset (DFO) pre-compensation to enhance spectrum utilization efficiency. By calculating the relative motion between the satellite and the transmitting terminal, combined with the length and transmission rate of the message, the optimal compensation value of the Doppler frequency deviation is determined. The frequency-protection interval is reduced. Simulation results show that the pre-compensation method can expand the user access volume by 90-400 times. Properly selecting the number of message splits and transmission rate to perform DFO pre-compensation calculations can increase user access by an additional 45% or more. This method improves the spectrum utilization efficiency and provides a solution to the challenge of access by a large number of users.
This article presents a comprehensive theoretical analysis of the resilience demonstrated by the three-stage Doherty power amplifier (DPA) when operating under load mismatch conditions. Additionally, a novel reconfigurable three-stage DPA architecture is introduced, with the aim of enhancing resilience to load mismatch using exceptionally simple circuits and a one-dimensional (1D) control method. To validate the efficacy of this proposed architecture and control approach, a DPA prototype employing commercial gallium nitride (GaN) active devices has been designed and meticulously fabricated at 2 GHz. With a matched 50 Ω load, the fabricated three-stage DPA achieves a high-efficiency range of 9.5 dB with larger than 51% back-off drain efficiency (DE). Through the proposed 1D control, the DPA presents 47.0%-55.1% back-off efficiency with ≤ 2 dB power fluctuation at a 2:1 voltage standing wave ratio (VSWR) over a 360° phase span. When driven by a 20 MHz long-term evolution (LTE) signal with an 8 dB peak-to-average power ratio (PAPR), the DPA achieves 46.2%-53.9% average efficiency and better than -21 dBc adjacent channel power ratio (ACPR) without digital pre-distortion (DPD) under load mismatch conditions.