Wildlife ecologists and biologists recapture deer to collect tracking data from deer collars or wait for a drop-off of a deer collar construction that is automatically detached and disconnected. The research teams need to manage a base camp with medical trailers, helicopters, and airplanes to capture deer or wait for several months until the deer collar drops off of the deer's neck. We propose an intelligent robust base-station research with a low-cost and time saving method to obtain recording sensor data from their collars to a listener node, and readings are obtained without opening the weatherproof deer collar. We successfully designed the and implemented a robust base station system for automatically collecting data of the collars and listener motes in harsh wilderness environments. Intelligent solutions were also analyzed for improved data collections and pattern predictions with drone-based detection and tracking algorithms.

Recurrent neural network (RNN) architectures have been used for language modeling (LM) tasks that require learning long-range word or character sequences. However, the RNN architecture is still suffered from unstable gradients on long-range sequences. To address the issue of long-range sequences, an attention mechanism has been used, showing state-of-the-art (SOTA) performance in all LM tasks. A differentiable neural computer (DNC) is a deep learning architecture using an attention mechanism. The DNC architecture is a neural network augmented with a content-addressable external memory. However, in the write operation, some information unrelated to the input word remains in memory. Moreover, DNCs have been found to perform poorly with low numbers of weight parameters. Therefore, we propose a robust memory deallocation method using a limited retention vector. The limited retention vector determines whether the network increases or decreases its usage of information in external memory according to a threshold. We experimentally evaluate the robustness of a DNC implementing the proposed approach according to the size of the controller and external memory on the enwik8 LM task. When we decreased the number of weight parameters by 32.47%, the proposed DNC showed a low bits-per-character (BPC) degradation of 4.30%, demonstrating the effectiveness of our approach in language modeling tasks.

In recent years, deep reinforcement learning (DRL) models are enjoying great interest as their success in a variety of challenging tasks. Deep Q-Network (DQN) is a widely used deep reinforcement learning model, which trains an intelligent agent that executes optimal actions while interacting with an environment. This model is well known for its ability to surpass skilled human players across many Atari 2600 games. Although DQN has achieved excellent performance in practice, there lacks a clear understanding of why the model works. In this paper, we present a visual analytics system for understanding deep Q-network in a non-blind matter. Based on the stored data generated from the training and testing process, four coordinated views are designed to expose the internal execution mechanism of DQN from different perspectives. We report the system performance and demonstrate its effectiveness through two case studies. By using our system, users can learn the relationship between states and Q-values, the function of convolutional layers, the strategies learned by DQN and the rationality of decisions made by the agent.

With the widespread deployment of the fifth-generation (5G) communication networks, various real-time applications are rapidly increasing and generating massive traffic on backhaul network environments. In this scenario, network congestion will occur when the communication and computation resources exceed the maximum available capacity, which severely degrades the network performance. To alleviate this problem, this paper proposed an intelligent resource allocation (IRA) to integrate with the extant resource adjustment (ERA) approach mainly based on the convergence of support vector machine (SVM) algorithm, software-defined networking (SDN), and mobile edge computing (MEC) paradigms. The proposed scheme acquires predictable schedules to adapt the downlink (DL) transmission towards off-peak hour intervals as a predominant priority. Accordingly, the peak hour bandwidth resources for serving real-time uplink (UL) transmission enlarge its capacity for a variety of mission-critical applications. Furthermore, to advance and boost gateway computation resources, MEC servers are implemented and integrated with the proposed scheme in this study. In the conclusive simulation results, the performance evaluation analyzes and compares the proposed scheme with the conventional approach over a variety of QoS metrics including network delay, jitter, packet drop ratio, packet delivery ratio, and throughput.

Previous QoS-aware service composition methods mainly focus on how to generate composite service with the optimal QoS efficiently for a single request. However, in the real application scenarios, there are multiple service requests and multiple service providers. It is more important to compose services with suboptimal QoS and maintain the load balance between services. To solve this problem, in this paper, we propose a service composition method, named as dynamically change and balancing composition method (DCBC). It assumes that the QoS of service is not static, and the services can adjust the value of QoS to gain more opportunities to be selected for composition. The method mainly includes two steps, which are the preprocessing step and the service selection step. In the preprocessing step, a backward global best QoS calculation is performed which regarding the static and dynamic QoS respectively; then guided by the global QoS, the feasible services can be selected efficiently in the service selection step. The experiments show that the DCBC method can not only improve the overall quality of composite services but also guarantee the fulfill ratio of requests and the load balance of services.

Recently, most cloud services use Docker container environment to provide their services. However, there are no researches to evaluate the performance of communication libraries for multi-GPU based distributed deep learning in a Docker container environment. In this paper, we propose an efficient communication architecture for multi-GPU based deep learning in a Docker container environment by evaluating the performances of various communication libraries. We compare the performances of the parameter server architecture and the All-reduce architecture, which are typical distributed deep learning architectures. Further, we analyze the performances of two separate multi-GPU resource allocation policies - allocating a single GPU to each Docker container and allocating multiple GPUs to each Docker container. We also experiment with the scalability of collective communication by increasing the number of GPUs from one to four. Through experiments, we compare OpenMPI and MPICH, which are representative open source MPI libraries, and NCCL, which is NVIDIA's collective communication library for the multi-GPU setting. In the parameter server architecture, we show that using CUDA-aware OpenMPI with multi-GPU per Docker container environment reduces communication latency by up to 75%. Also, we show that using NCCL in All-reduce architecture reduces communication latency by up to 93% compared to other libraries.

The effectiveness of testing in Model-based Testing (MBT) for Software Product Line (SPL) can be achieved by considering fault detection in test case. The lack of fault consideration caused test case in test suite to be listed randomly. Test Case Prioritization (TCP) is one of regression techniques that is adaptively capable to detect faults as early as possible by reordering test cases based on fault detection rate. However, there is a lack of studies that measured faults in MBT for SPL. This paper proposes a Test Case Prioritization (TCP) approach based on dissimilarity and string based distance called Last Minimal for Local Maximal Distance (LM-LMD) with Dice-Jaro-Winkler Dissimilarity. LM-LMD with Dice-Jaro-Winkler Dissimilarity adopts Local Maximum Distance as the prioritization algorithm and Dice-Jaro-Winkler similarity measure to evaluate distance among test cases. This work is based on the test case generated from statechart in Software Product Line (SPL) domain context. Our results are promising as LM-LMD with Dice-Jaro-Winkler Dissimilarity outperformed the original Local Maximum Distance, Global Maximum Distance and Enhanced All-yes Configuration algorithm in terms of Average Fault Detection Rate (APFD) and average prioritization time.

In recent years, mobile devices have become an essential part of daily life. More and more applications are being supported by mobile devices thanks to edge computing, which represents an emergent architecture that provides computing, storage, and networking capabilities for mobile devices. In edge computing, heavy tasks are offloaded to edge nodes to alleviate the computations on the mobile side. However, offloading computational tasks may incur extra energy consumption and delays due to network congestion and server queues. Therefore, it is necessary to optimize offloading decisions to minimize time, energy, and payment costs. In this article, different offloading models are examined to identify the offloading parameters that need to be optimized. The paper investigates and compares several optimization techniques used to optimize offloading decisions, specifically Swarm Intelligence (SI) models, since they are best suited to the distributed aspect of edge computing. Furthermore, based on the literature review, this study concludes that a Cuckoo Search Algorithm (CSA) in an edge-based architecture is a good solution for balancing energy consumption, time, and cost.

Recently, the healthcare field has undergone rapid changes owing to the accumulation of health big data and the development of machine learning. Data mining research in the field of healthcare has different characteristics from those of other data analyses, such as the structural complexity of the medical data, requirement for medical expertise, and security of personal medical information. Various methods have been implemented to address these issues, including the machine learning model and cloud platform. However, the machine learning model presents the problem of opaque result interpretation, and the cloud platform requires more in-depth research on security and efficiency. To address these issues, this paper presents a recent technology for Internet-of-Things-based (IoT-based) health big data processing. We present a cloud-based IoT health platform and health big data processing technology that reduces the medical data management costs and enhances safety. We also present a data mining technology for health-risk prediction, which is the core of healthcare. Finally, we propose a study using explainable artificial intelligence that enhances the reliability and transparency of the decision-making system, which is called the black box model owing to its lack of transparency.

This paper proposes a new image retrieval method called the 3D integrated multi-index to fuse SIFT (Scale Invariant Feature Transform) visual words with other features at the indexing level. The advantage of the 3D integrated multi-index is that it can produce finer subdivisions in the search space. Compared with the inverted indices of medium-sized codebook, the proposed method increases time slightly in preprocessing and querying. Particularly, the SIFT, contour and colour features are fused into the integrated multi-index, and the joint cooperation of complementary features significantly reduces the impact of false positive matches, so that effective image retrieval can be achieved. Extensive experiments on five benchmark datasets show that the 3D integrated multi-index significantly improves the retrieval accuracy. While compared with other methods, it requires an acceptable memory usage and query time. Importantly, we show that the 3D integrated multi-index is well complementary to many prior techniques, which make our method compared favorably with the state-of-the-arts.

Modern technology offers many ways to enhance teaching and learning that in turn promote the development of tools for educational activities both inside and outside the classroom. Many educational programs using the augmented reality (AR) technology are being widely used to provide supplementary learning materials for students. This paper describes the potential and challenges of using GeoGebra AR in mathematical studies, whereby students can view 3D geometric objects for a better understanding of their structure, and verifies the feasibility of its use based on experimental results. The GeoGebra software can be used to draw geometric objects, and 3D geometric objects can be viewed using AR software or AR applications on mobile phones or computer tablets. These could provide some of the required materials for mathematical education at high schools or universities. The use of the GeoGebra application for education in Laos will be particularly discussed in this paper.

The edges with "bridge characteristic" play the role of connecting the communication between regions in power optical cable network. To solve the problem of mining edges with "bridge characteristic" in provincial power optical cable network, the complex power optical cable network model is constructed. Firstly, to measure the generated potential energy of all nodes in n-level neighborhood local structure for one edge, the n-level neighborhood local structure topological potential is designed. And the minimum connected chain attenuation is designed to measure the attenuation degree caused by substituted edges. On the basis of that, the minimum connected chain attenuation topological potential based measurement is designed. By using the designed measurement, a bridge-edges mining algorithm is proposed to mine edges with "bridge characteristic". The experiments are conducted on the physical topology of the power optical cable network in Jilin Province. Compared with that of other three typical methods, the network efficiency and connectivity of the proposed method are decreased by 3.58% and 28.79% on average respectively. And the proposed method can not only mine optical cable connection with typical "bridge characteristic" but also can mine optical cables without obvious characteristics of city or voltage, but it have "bridge characteristic" in the topology structure.

The secure communication of information is a major concern over the internet. The information must be protected before transmitting over a communication channel to avoid security violations. In this paper, a new hybrid method called compressed encrypted data embedding (CEDE) is proposed. In CEDE, the secret information is first compressed with Lempel Ziv Welch (LZW) compression algorithm. Then, the compressed secret information is encrypted using the Advanced Encryption Standard (AES) symmetric block cipher. In the last step, the encrypted information is embedded into an image of size 512 × 512 pixels by using image steganography. In the steganographic technique, the compressed and encrypted secret data bits are divided into pairs of two bits and pixels of the cover image are also arranged in four pairs. The four pairs of secret data are compared with the respective four pairs of each cover pixel which leads to sixteen possibilities of matching in between secret data pairs and pairs of cover pixels. The least significant bits (LSBs) of current and imminent pixels are modified according to the matching case number. The proposed technique provides double-folded security and the results show that stego image carries a high capacity of secret data with adequate peak signal to noise ratio (PSNR) and lower mean square error (MSE) when compared with existing methods in the literature.

Traditional image steganography hides secret information by embedding, which inevitably leaves modification traces and is easy to be detected by steganography analysis tools. Since coverless steganography can effectively resist steganalysis, it has become a hotspot in information hiding research recently. Most coverless image steganography (CIS) methods are based on mapping rules, which not only exposes the vulnerability to geometric attacks, but also are less secure due to the revelation of mapping rules. To address the above issues, we introduced camouflage images for steganography instead of directly sending stego-image, which further improves the security performance and information hiding ability of steganography scheme. In particular, based on the different sub-features of stego-image and potential camouflage images, we try to find a larger similarity between them so as to achieve the reversible steganography. Specifically, based on the existing CIS mapping algorithm, we first can establish the correlation between stego-image and secret information and then transmit the camouflage images, which are obtained by reversible sub-feature retrieval algorithm. The received camouflage image can be used to reverse retrieve the stego-image in a public image database. Finally, we can use the same mapping rules to restore secret information. Extensive experimental results demonstrate the better robustness and security of the proposed approach in comparison to state-of-art CIS methods, especially in the robustness of geometric attacks.

In a wide range of ML applications, the training data contains privacy-sensitive information that should be kept secure. Training the ML systems by privacy-sensitive data makes the ML model inherent to the data. As the structure of the model has been fine-tuned by training data, the model can be abused for accessing the data by the estimation in a reverse process called model inversion attack (MIA). Although, MIA has been applied to shallow neural network models of recognizers in literature and its threat in privacy violation has been approved, in the case of a deep learning (DL) model, its efficiency was under question. It was due to the complexity of a DL model structure, big number of DL model parameters, the huge size of training data, big number of registered users to a DL model and thereof big number of class labels. This research work first analyses the possibility of MIA on a deep learning model of a recognition system, namely a face recognizer. Second, despite the conventional MIA under the white box scenario of having partial access to the users' non-sensitive information in addition to the model structure, the MIA is implemented on a deep face recognition system by just having the model structure and parameters but not any user information. In this aspect, it is under a semi-white box scenario or in other words a gray-box scenario. The experimental results in targeting five registered users of a CNN-based face recognition system approve the possibility of regeneration of users' face images even for a deep model by MIA under a gray box scenario. Although, for some images the evaluation recognition score is low and the generated images are not easily recognizable, but for some other images the score is high and facial features of the targeted identities are observable. The objective and subjective evaluations demonstrate that privacy cyber-attack by MIA on a deep recognition system not only is feasible but also is a serious threat with increasing alert state in the future as there is considerable potential for integration more advanced ML techniques to MIA.

Among the scientific literature, it has not been possible to find a consensus on the definition of the limits or properties that allow differentiating or grouping the cyber-physical systems (CPS) and the Internet of Things (IoT). Despite this controversy the papers reviewed agree that both have become crucial elements not only for industry but also for society in general. The impact of a malware attack affecting one of these systems may suppose a risk for the industrial processes involved and perhaps also for society in general if the system affected is a critical infrastructure. This article reviews the state of the art of the application of machine learning in the automation of malware detection in cyberphysical systems, evaluating the most representative articles in this field and summarizing the results obtained, the most common malware attacks in this type of systems, the most promising algorithms for malware detection in cyberphysical systems and the future lines of research in this field with the greatest potential for the coming years.

Up to now, many state-of-arts have been proposed for two-way relaying system with linear modulations. The performances of antenna selection (AS) at both transmit and relay nodes need to be investigated in some two-way relaying multiple-input multiple-output (TWRM) systems. In this paper, the goal is focused on the study of nonlinear modulations, i.e., continuous phase modulation (CPM) in TWRM systems. Firstly, the joint phase trellis are simplified by reversed Rimoldi processing so as to reduce the systems' complexity. Then the performances of joint transmit and receive antenna selection (JTRAS) with CPM modulations in two-way relaying MIMO systems are analyzed. More exactly, the pair wise probability (PEP) is used to evaluate the error performance based on the CPM signal matrix, which is calculated in terms of Laurent expression. Since the channels subject to two terminal nodes share common antennas at relay node R in multiple-access scheme, we revise the JTRAS algorithm and compare it to existing algorithm via simulation. Finally, the error performances for various schemes of antenna selection are simulated and compared to the analysis in this paper.

In this paper, we propose a new joint precoder and postcoder design strategy to support multiple streams per user in multiuser multiple-input multiple-output (MIMO) systems. We propose two step precoding strategies using equal channel gain decomposition and block diagonalization at the transmitter. With the proposed precoder, the multiuser MIMO channel can be decomposed into multiple parallel channels with equal channel gain per user. After applying receive postcoder which is generated and sent by the transmitter, we can use ML based decoder per stream to achieve full receive diversity. Achievable sum rate bound and diversity performance of the proposed algorithm are presented with feedback signaling design and quantitative complexity analysis. Simulation results show that the proposed algorithm asymptotically approaches to the sum rate capacity of the MIMO broadcast channel while maintaining full diversity order.

With the assistance of non-orthogonal multiple access (NOMA), the spectrum efficiency and the number of users in cloud radio access network (CRAN) can be greatly improved. In this paper, the system performance of NOMA-assisted CRAN is investigated. Specially, the outage probability (OP) and ergodic sum rate (ESR), are derived for performance evaluation of the system, respectively. Based on this, by minimizing the OP of the system, a suboptimal power allocation (PA) scheme with closed-form PA coefficients is proposed. Numerical simulations validate the accuracy of the theoretical results, where the derived OP has more accuracy than the existing one. Moreover, the developed PA scheme has superior performance over the conventional fixed PA scheme but has smaller performance loss than the optimal PA scheme using the exhaustive search method.