Clothes-Changing person re-identification algorithm (Re-ID) is the task of retrieving the query person in the case of the change of pedestrian clothing. Changes in pedestrian clothing lead to an offset in clothing features, resulting in a decrease in identification performance. Simply removing clothing may lead to the loss of contour information. Furthermore, algorithms based on feature decoupling cannot guarantee the accuracy of the positional relationship between clothing features and other features due to the lack of groundtruth. To address these issues, we propose a novel clothes-changing person re-identification algorithm based on causal feature supervision decoupling. Utilizing a multi-scale feature fusion module to extract fine-grained features of clothing and add supervised information labels. This enables the dual-branch network to separately approach overall and clothing features, promoting the extraction of effective identity information by the causal decoupling module, and obtaining unbiased estimations of pedestrians. Experimental results show that the proposed algorithm achieves the highest mAP and Top-1 accuracy on the LTCC and PRCC datasets. The source code is available at https://github.com/zhihu250/CISupNet.

Radio Frequency Fingerprint (RFF) based physical layer authentication technology provides enhanced security for wireless communications. However, the spatiotemporal overlap of wireless signals makes it challenging to label wireless device samples. Moreover, generative networks, such as Generative Adversarial Networks (GANs) struggle to retain the subtle signal features. Utilizing existing unlabeled samples poses a significant challenge for RFF data augmentation. This paper proposes the RF Distillation Diffusion (RFDD) model, an efficient method for RFF data augmentation. RFDD employs a conditional diffusion model to generate high-quality RF signals, which fully utilizes existing unlabeled samples to learn the data distribution of signals, and labeled samples are used to enhance the capability of RFF feature extraction. Additionally, knowledge distillation is utilized to improve sampling efficiency. Experimental results show that the RFDD can fully use 20% unlabeled samples to generate high-quality synthetic RF signals within only 0.45 seconds and improve RFF identification accuracy by 25.63% with 40% synthetic signals under SNRs from -5 to 5 dB. The code is available at https://github.com/XMU-Kai/RFDD

ABSTRACT Automatic modulation recognition enables rapid spectrum access and serves as a key technical component for achieving communication‐aware integration. In recent years, deep learning methods have attracted considerable attention in this field. However, their performance relies heavily on the availability of large‐scale datasets. Data augmentation has proven effective in mitigating data scarcity. To address this challenge, this paper proposes a data augmentation algorithm based on a conditional diffusion model to improve model training under limited data conditions. In the proposed framework, the noise observation model detects noise in the input signal at different time steps. The identified noise is then iteratively removed during the reverse process of the conditional diffusion model to generate the corresponding modulated signals. Generated signals with high confidence are incorporated into the training set to enhance data diversity. Experimental results demonstrate that the proposed algorithm significantly improves the classification network's performance and outperforms existing data augmentation approaches.

Radio frequency (RF) fingerprint recognition uses the hardware characteristic differences of wireless devices to achieve identity authentication from the physical layer. However, there are significant differences in the distribution of the RF fingerprint feature of the same transmitter signal collected in different receiving devices. The existing deep learning algorithms usually retrain the model to improve the recognition accuracy, which brings challenges such as the high cost of retraining and a lack of training samples. Transfer learning can promote the learning of new tasks by extracting knowledge from existing tasks, and can be used to reduce feature distribution differences between different data sets. In this paper, we optimize the CycleGAN network, which is widely used in transfer learning, and propose a Dual Transfer-Generative Adversarial Network (DTGAN). This network replaces the transposed convolution with one-dimensional linear interpolation, which better aligns the feature boundaries of the target domain signal with the source domain signal. The experimental results show that without retraining the model, the recognition accuracy of data after feature transfer is significantly better than that of the non-transfer scenario, which enhances the robustness of RF fingerprint recognition under different receiving conditions.

Most methods rely on unsupervised learning due to the limited availability of anomaly data. However, most of the current unsupervised learning methods are based on deep self-encoders, which do not pay enough attention to the consistency of the motion process. Therefore, we propose a video anomaly detection framework based on motion consistency (VADMC).The framework uses a CVAE network as a generator to generate predicted frames. In order to increase the reconstruction error of the CVAE network, we embed a memory module in the optical flow coding features, which is used to memorize the feature distribution of the normal patterns. The reconstruction error is increased by perturbing the a priori distribution, thus increasing the reconstruction error. A discriminator is used to discriminate the generated optical flow maps to ensure the consistency of the forward and backward motions of the normal samples. We conducted experiments on three public datasets to demonstrate the effectiveness of the VADMC framework. The accuracy on the UCSD PED2, CHUK Avenue, and Shanghai Tech datasets reached 97.2%, 76.3%, and 76.2%, respectively. Compared with previous state-of-the-art methods, our method shows competitive results.

Automatic modulation recognition techniques enable fast spectrum access. In recent years, deep learning methods have received much attention in the field of modulation recognition. However, its performance requires a large number of data samples as support. Data augmentation methods are regarded as one of the effective ways to solve the insufficient data samples. To this end, this paper proposes a data augmentation algorithm based on the conditional diffusion model to solve the problem that the model cannot be adequately trained in the case of insufficient data. In the proposed algorithm, the noise observation model acquires the noise in the input signal at different time steps. Then, the input is continuously denoised by removing the noise observed by the noise observation model during the inverse process of the conditional diffusion model to generate the corresponding modulated signal. We put the modulation modes with high confidence in the generated signals into the training set for data augmentation. Experimental results show that the data augmentation algorithm based on the conditional diffusion model proposed in this paper can effectively improve the classification performance of the model.

Haze can reduce the visibility of the captured image, making it hard to accurately distinguish the details of each object in the captured image scene. Aiming at the problem of detail loss in existing dehazing models, this paper proposes a lightweight end-to-end image dehazing framework called DFE-GAN (Detail Feature Enhancement-GAN). The missing detail contours in the haze image can be predicted by employing a densely connected detail feature prediction network. Supplemented with a patch discriminator and an improved loss function, the restoration of details in the dehazing image is enhanced to improve image quality. We apply inverse residual modules to extract and fuse multi-scale features from images, which can ensure the real-time processing capability of the model. Compared with previous state-of-the-art approaches, solid experimental results on various benchmark datasets validate the robustness and effectiveness of our model.

Due to the limited availability of anomalous samples for training, video anomaly detection is commonly viewed as a one-class classification problem. Many prevalent methods investigate the reconstruction difference produced by AutoEncoders (AEs) under the assumption that the AEs would reconstruct the normal data well while reconstructing anomalies poorly. However, even with only normal data training, the AEs often reconstruct anomalies well, which depletes their anomaly detection performance. To alleviate this issue, we propose a simple yet efficient framework for video anomaly detection. The pseudo anomaly samples are introduced, which are synthesized from only normal data by embedding random mask tokens without extra data processing. We also propose a normalcy consistency training strategy that encourages the AEs to better learn the regular knowledge from normal and corresponding pseudo anomaly data. This way, the AEs learn more distinct reconstruction boundaries between normal and abnormal data, resulting in superior anomaly discrimination capability. Experimental results demonstrate the effectiveness of the proposed method.

Video anomaly detection is an essential but challenging task. The prevalent methods mainly investigate the reconstruction difference between normal and abnormal patterns but ignore the semantics consistency between appearance and motion information of behavior patterns, making the results highly dependent on the local context of frame sequences and lacking the understanding of behavior semantics. To address this issue, we propose a framework of Appearance-Motion Semantics Representation Consistency that uses the gap of appearance and motion semantic representation consistency between normal and abnormal data. The two-stream structure is designed to encode the appearance and motion information representation of normal samples, and a novel consistency loss is proposed to enhance the consistency of feature semantics so that anomalies with low consistency can be identified. Moreover, the lower consistency features of anomalies can be used to deteriorate the quality of the predicted frame, which makes anomalies easier to spot. Experimental results demonstrate the effectiveness of the proposed method.

Frame prediction based on AutoEncoder plays a significant role in unsupervised video anomaly detection. Ideally, the models trained on the normal data could generate larger prediction errors of anomalies. However, the correlation between appearance and motion information is underutilized, which makes the models lack an understanding of normal patterns. Moreover, the models do not work well due to the uncontrollable generalizability of deep AutoEncoder. To tackle these problems, we propose a multi-level memory-augmented appearance-motion correspondence framework. The latent correspondence between appearance and motion is explored via appearance-motion semantics alignment and semantics replacement training. Besides, we also introduce a Memory-Guided Suppression Module, which utilizes the difference from normal prototype features to suppress the reconstruction capacity caused by skip-connection, achieving the tradeoff between the good reconstruction of normal data and the poor reconstruction of abnormal data. Experimental results show that our framework outperforms the state-of-the-art methods, achieving AUCs of 99.6%, 93.8%, and 76.3% on UCSD Ped2, CUHK Avenue, and ShanghaiTech datasets.

Radio Frequency Fingerprint (RFF) technology is an effective means to defend against cheating and counterfeiting attacks in wireless communication. However, to move from a theoretical algorithm to a practical application, the challenges of imbalanced data samples and environmental noise must be addressed for Radio Frequency (RF) identification technology. Although noise reduction can restore the signal to some extent, the recognition performance of RFF technology is affected when the dataset is imbalanced. While many RF identification algorithms focus on identification performance under a low Signal-to-Noise Ratio (SNR), performance degradation caused by data imbalance is a pressing problem that requires attention. Directly applying re-sampling algorithms in imbalanced dataset processing can lead to data overlap and neural network over-fitting. To address these issues, this paper proposes a “Redistillation of Radio Frequency Knowledge” (RRFK) algorithm combined with knowledge distillation (KD). The experimental results show that the proposed algorithm can achieve good recognition performance in both stepped and long-tail imbalanced data sets.

Radio Frequency Fingerprint (RFF) technology is an effective means to defend against cheating and counterfeit attacks in wireless communication. A deep learning-based RFF recognition algorithm can achieve well recognition performance, but it needs many balanced samples to train the model. However, the problem of sample imbalance is widespread in RFF identification tasks, and the number of signal samples of illegal devices is minimal. Neural network models usually can't learn these minority representations well, which seriously affects the performance of RFF recognition. Many advanced algorithms proposed to alleviate the problem of data imbalance don't perform well in the task of RFF recognition because they ignore the characteristics of RFF signals. Therefore, an algorithm based on Random Railings Enhancement (RRE) is proposed in this paper, which fills the data set with random masks according to the signal values front and rear. RRE protects the original signal's information, effectively expands the rare dataset, and has the effect of data enhancement. The experimental results show that the RRE can improve the performance of Radio Frequency (RF) identification technology tasks in the case of imbalanced data sets.

Underwater image enhancement algorithms improve image quality and indirectly enhance underwater visibility. Although many underwater image enhancement neural networks have been proposed, they require large amounts of data. To reduce the amount of data required while providing better image enhancement, this study proposes an underwater image colour transfer generative adversarial network (UCT‐GAN). The authors first design a non‐linear mapping function to generate colour cast images according to original images. Then, the authors utilise these image pairs (i.e. colour cast images and corresponding original images) to guide the UCT‐GAN in learning the inverse function of the designed non‐linear mapping function. Finally, colour cast images are restored via the inverse function. A data augmentation method based on Poisson fusion and block combination is also proposed to overcome the problem of requiring a large amount of training data. Moreover, the authors extend UCT‐GAN into a multi‐class colour transfer network to achieve an array of underwater image enhancements. Experimental results indicate that the proposed UCT‐GAN can more effectively resolve underwater image colour cast compared to existing algorithms.

With the development of artificial intelligence and self-driving, vehicular ad-hoc network (VANET) has become an irreplaceable part of the Intelligent Transportation Systems (ITSs). However, the traditional network of the ground cannot meet the requirements of transmission, processing, and storage among vehicles. Under this circumstance, integrating space and air nodes into the whole network can provide comprehensive traffic information and reduce the transmission delay. The high mobility and low latency in the Space-Air-Ground Integrated Network (SAGIN) put forward higher requirements for security issues such as identity authentication, privacy protection and data security. This paper simplifies the Blockchain and proposes an identity authentication and privacy protection scheme based on the Hashchain in the SAGIN. The scheme focuses on the characteristics of the wireless signal to identify and authenticate the nodes. The verification and backup of the records on the block are implemented with the distributed streaming platform, Kafka algorithm, instead of the consensus. Furthermore, this paper analyzes the security of this scheme. Afterward, the experimental results reveal the delay brought by the scheme using the simulation of SUMO, OMNeT++, and Veins.

Beyond their benign uses, the small Unmanned Aerial Vehicles (UAVs) are expected to take the major role in future smart cities that have attracted the attention of the public and authorities. Therefore, detecting, tracking and classifying the type of UAVs is important for surveillance and air traffic management applications. Existing UAVs detection works focus on radars, visual detection, and acoustic sensors. However, the work was done by applying Support Vector Machine (SVM), k-Nearest Neighbor (KNN) based methods to classify the UAVs need a large number of samples for feature extraction to train a model. In this paper, we propose a new small UAVs classification system using Auxiliary Classifier Wasserstein Generative Adversarial Networks (AC-WGANs) based on the wireless signals collected from the UAVs of various types. Before the classification, using the Universal Software Radio Peripheral (USRP), oscilloscope and antenna to collect the wireless signals, preprocessing and dimensionality reduction to represent information at a lower dimension space. The processed data from UAVs is input to the UAVs' discriminant model of the AC-WGANs for classification. The obtained results show the effectiveness of the proposed system, which can achieve a recognition accuracy of around 95% in the indoor environment and can also be suitable in the outdoor environment.

With the rapid development of science and technology, unmanned aerial vehicles (UAVs) gradually become the worldwide focus of science and technology. Not only the development and application but also the security of UAV is of great significance to modern society. Different from methods using radar, optical or acoustic sensors to detect UAV, this paper proposes a novel distance-based support vector data description (SVDD) algorithm using hash fingerprint as feature. This algorithm does not need large number of training samples and its computation complexity is low. Hash fingerprint is generated by extracting features of signal preamble waveforms. Distance-based SVDD algorithm is employed to efficiently detect and recognize low, slow, small unmanned aerial vehicles (LSSUAVs) using 2.4GHz frequency band.

This paper proposes a full-duplex physical security model with self-interference remaining. This model doesn't need the assistance of external jamming nodes and it ensures that the uplink and downlink transmission of the full-duplex system can achieve the required secrecy rate. Meanwhile, it creates a base station transmission power optimization problem with flexible constraints and a two-level method to achieve the optimization, so that power optimization can take place in the model with self-interference remaining.

ABSTRACT This paper proposes a time–frequency‐based fingerprinting identification approach by extracting the transient characterizations regarding highly integrated wireless local area network (WLAN) cards. The transient energy envelope is derived from the slice of spectrogram in the time–frequency domain. Then this transient response is fitted to a polynomial under least square criteria, and the polynomial coefficients are regarded as the feature vector. A data acquisition system has been set up to capture IEEE 802.11b Wi‐Fi (wireless fidelity) signals. The results exhibit an approving distinctiveness of up to 94% to classify different manufactories of WLAN cards and of 78.4% for the WLAN cards of the same manufactory. Copyright © 2011 John Wiley & Sons, Ltd.