The Generative Adversarial Networks, as one of the most creative deep learning models in recent years, has achieved great success in computer vision and natural language processing. It uses the game theory to generate the best sample in generator and discriminator. Recently, many deep learning models have been applied to the security field. Along with the idea of "generative" and "adversarial", researchers are trying to apply Generative Adversarial Networks to the security field. This paper presents the development of Generative Adversarial Networks. We review traditional generation models and typical Generative Adversarial Networks models, analyze the application of their models in natural language processing and computer vision. To emphasize that Generative Adversarial Networks models are feasible to be used in security, we separately review the contributions that their defenses in information security, cyber security and artificial intelligence security. Finally, drawing on the reviewed literature, we provide a broader outlook of this research direction.

This paper proposes the Lempel-Ziv 4(LZ4) compression accelerator optimized for scale-out servers in data centers. In order to reduce CPU loads caused by compression, we propose an accelerator solution and implement the accelerator on an Field Programmable Gate Array(FPGA) as heterogeneous computing. The LZ4 compression hardware accelerator is a fully pipelined architecture and applies 16 dictionaries to enhance the parallelism for high throughput compressor. Our hardware accelerator is based on the 20-stage pipeline and dictionary architecture, highly customized to LZ4 compression algorithm and parallel hardware implementation. Proposing dictionary architecture allows achieving high throughput by comparing input sequences in multiple dictionaries simultaneously compared to a single dictionary. The experimental results provide the high throughput with intensively optimized in the FPGA. Additionally, we compare our implementation to CPU implementation results of LZ4 to provide insights on FPGA-based data centers. The proposed accelerator achieves the compression throughput of 639MB/s with fine parallelism to be deployed into scale-out servers. This approach enables the low power Intel Atom processor to realize the Hadoop storage along with the compression accelerator.

Modulation recognition (MR) plays a key role in cognitive radar, cognitive radio, and some other civilian and military fields. While existing methods can identify the signal modulation type by extracting the signal characteristics, the quality of feature extraction has a serious impact on the recognition results. In this paper, an end-to-end MR method based on long short-term memory (LSTM) and the gated recurrent unit (GRU) is put forward, which can directly predict the modulation type from a sampled signal. Additionally, the sliding window method is applied to fast-changing mixed-modulation signals for which the signal modulation type changes over time. The recognition accuracy on training datasets in different SNR ranges and the proportion of each modulation method in misclassified samples are analyzed, and it is found to be reasonable to select the evenly-distributed and full range of SNR data as the training data. With the improvement of the SNR, the recognition accuracy increases rapidly. When the length of the training dataset increases, the neural network recognition effect is better. The loss function value of the neural network decreases with the increase of the training dataset length, and then tends to be stable. Moreover, when the fast-changing period is less than 20ms, the error rate is as high as 50%. As the fast-changing period is increased to 30ms, the error rates of the GRU and LSTM neural networks are less than 5%.

The Unmanned Aerial Vehicles (UAV) networks consisting of low-cost UAVs are very vulnerable to smart jammers that can choose their jamming policies based on the ongoing communication policies accordingly. In this article, we propose a novel cloud and edge-aided mobile communication scheme for low-cost UAV network against smart jamming. The challenge of this problem is to design a communication scheme that not only meets the requirements of defending against smart jamming attack, but also can be deployed on low-cost UAV platforms. In addition, related studies neglect the problem of decision-making algorithm failure caused by intermittent ground-to-air communication. In this scheme, we use the policy network deployed on the cloud and edge servers to generate an emergency policy tables, and regularly update the generated policy table to the UAVs to solve the decision-making problem when communications are interrupted. In the operation of this communication scheme, UAVs need to offload massive computing tasks to the cloud or the edge servers. In order to prevent these computing tasks from being offloaded to a single computing resource, we deployed a lightweight game algorithm to ensure that the three types of computing resources, namely local, edge and cloud, can maximize their effectiveness. The simulation results show that our communication scheme has only a small decrease in the SINR of UAVs network in the case of momentary communication interruption, and the SINR performance of our algorithm is higher than that of the original Q-learning algorithm.

With various structured data, such as the company size, loan balance, and savings accounts, the voice of customer (VOC), which is text data containing contact history and counseling details was analyzed in this study. To analyze unstructured data, the term frequency-inverse document frequency (TF-IDF) analysis, semantic network analysis, sentiment analysis, and a convolutional neural network (CNN) were implemented. A performance comparison of the models revealed that the predictive model using the CNN provided the best performance with regard to predictive power, followed by the model using the TF-IDF, and then the model using semantic network analysis. In particular, a character-level CNN and a word-level CNN were developed separately, and the character-level CNN exhibited better performance, according to an analysis for the Korean language. Moreover, a systematic selection model for optimal text mining techniques was proposed, suggesting which analytical technique is appropriate for analyzing text data depending on the context. This study also provides evidence that the results of previous studies, indicating that individual customers leave when their loyalty and switching cost are low, are also applicable to corporate customers and suggests that VOC data indicating customers' needs are very effective for predicting their behavior.

Software defined networking (SDN) is considered to be one of the most promising paradigms in the future. To solve the scalability and performance problem that a single and centralized controller suffers from, the distributed multi-controller architecture is adopted, thus forms multi-domain SDN. In a multi-domain SDN network, it is of great importance to ensure a reliable control plane. In this paper, we focus on the reliability problem of multi-domain SDN against controller failure from perspectives of backup controller deployment and controller replication. We firstly propose a placement algorithm for backup controllers, which considers both the reliability and the cost factors. Then a controller replication mechanism based on shared data storage is proposed to solve the inconsistency between the active and standby controllers. We also propose a shared data storage layout method that considers both reliability and performance. Besides, a fault recovery and repair process is designed based on the controller backup and shared data storage mechanism. Simulations show that our approach can recover and repair controller failure. Evaluation results also show that the proposed backup controller placement approach is more effective than other methods.

Recently, the COVID 19 pandemic has affected on our daily lives and society in many ways. Specifically, it has brought rapid changes in the working environment from office working to smart telecommuting. In addition, cloud computing technology and services not only provided ubiquitous access, but also led to a sharing of information, internal-external communication channels, telework, and innovative smart work for the business process. As a result, smart work services based on social cloud networking have spread to the public sector. However, existing academic research examining smart work merely remains to focus on the theoretical conceptualization or to deal with merely several examples of private views. Best practices of smart work services based on cloud computing technology in the public field rarely exists. Moreover, many studies have been differently measured the values of smart work for private and public sectors depending on organizational singularities. Therefore, the study aims to define new theoretical implications and to explore future business strategies and policy directions based on a technical working group's personal psychological subjectivity. The research applied Q methodology, and selected five public organizations in Korea, that they have adopted or currently plan to adopt some part of smart work services.

This research aims to develop a deep learning-based method that automatically detects and segments the defects on railroad surfaces to reduce the cost of visual inspection of the railroad. We developed our segmentation model by modifying a fully convolutional network model [1], a well-known segmentation model used for machine learning, to detect and segment railroad surface defects. The data used in this research are images of the railroad surface with one or more defect regions. Railroad images were cropped to a suitable size, considering the long height and relatively narrow width of the images. They were also normalized based on the variance and mean of the data images. Using these images, the suggested model was trained to segment the defect regions. The proposed method showed promising results in the segmentation of defects. We consider that the proposed method can facilitate decision-making about railroad maintenance, and potentially be applied for other analyses.

We propose a rapid adaptive learning framework for streaming object detection, called EER-ASSL. The method combines the expected error reduction (EER) dependent rollback learning and the active semi-supervised learning (ASSL) for a rapid adaptive CNN detector. Most CNN object detectors are built on the assumption of static data distribution. However, images are often noisy and biased, and the data distribution is imbalanced in a real world environment. The proposed method consists of collaborative sampling and EER-ASSL. The EER-ASSL utilizes the active learning (AL) and rollback based semi-supervised learning (SSL). The AL allows us to select more informative and representative samples measuring uncertainty and diversity. The SSL divides the selected streaming image samples into the bins and each bin repeatedly transfers the discriminative knowledge of the EER and CNN models to the next bin until convergence and incorporation with the EER rollback learning algorithm is achieved. The EER models provide a rapid short-term myopic adaptation and the CNN models an incremental long-term performance improvement. EER-ASSL can overcome noisy and biased labels in varying data distribution. Extensive experiments shows that EER-ASSL obtained 70.9 mAP compared to state-of-the-art technology such as Faster RCNN, SSD300, and YOLOv2.

Drone detection can be considered as a specific sort of small object detection, which has always been a challenge because of its small size and few features. For improving the detection rate of drones, we design a Deeper SSD network, which uses large-scale input image and deeper convolutional network to obtain more features that benefit small object classification. At the same time, in order to improve object classification performance, we implemented the up-sampling modules to increase the number of features for the low-level feature map. In addition, in order to improve object location performance, we adopted the down-sampling modules so that the context information can be used by the high-level feature map directly. Our proposed Deeper SSD and its variants are successfully applied to the self-designed drone datasets. Our experiments demonstrate the effectiveness of the Deeper SSD and its variants, which are useful to small drone's detection and recognition. These proposed methods can also detect small and large objects simultaneously.

This paper proposes transfer learning and fine-tuning techniques for a deep learning model to detect three distinct brain tumors from Magnetic Resonance Imaging (MRI) scans. In this work, the recent YOLOv4 model trained using a collection of 3064 T1-weighted Contrast-Enhanced (CE)-MRI scans that were pre-processed and labeled for the task. This work trained with the partial 29-layer YOLOv4-Tiny and fine-tuned to work optimally and run efficiently in most platforms with reliable performance. With the help of transfer learning, the model had initial leverage to train faster with pre-trained weights from the COCO dataset, generating a robust set of features required for brain tumor detection. The results yielded the highest mean average precision of 93.14%, a 90.34% precision, 88.58% recall, and 89.45% F1-Score outperforming other previous versions of the YOLO detection models and other studies that used bounding box detections for the same task like Faster R-CNN. As concluded, the YOLOv4-Tiny can work efficiently to detect brain tumors automatically at a rapid phase with the help of proper fine-tuning and transfer learning. This work contributes mainly to assist medical experts in the diagnostic process of brain tumors.

The recent expansion of the digital content market is increasing the technical demand for various facial image transformations within the virtual environment. The recent image translation technology enables changes between various domains. However, current image-to-image translation techniques do not provide stable performance through unsupervised learning, especially for shape learning in the face transition field. This is because the face is a highly sensitive feature, and the quality of the resulting image is significantly affected, especially if the transitions in the eyes, nose, and mouth are not effectively performed. We herein propose a new unsupervised method that can transform an in-wild face image into another face style through radical transformation. Specifically, the proposed method applies two face-specific feature loss functions for a generative adversarial network. The proposed technique shows that stable domain conversion to other domains is possible while maintaining the image characteristics in the eyes, nose, and mouth.

The traditional roles of leaders are to influence members and motivate them to achieve shared goals in organizations. However, leaders such as top managers and chief executive officers, in practice, do not always directly meet or influence other company members. In fact, they tend to have the greatest impact on their members through formal speeches, company procedures, and the like. As such, official speech is directly related to the motivation of company employees. In an official speech, not only the contents of the speech, but also the voice characteristics of the speaker have an important influence on listeners, as the different vocal characteristics of a person can have different effects on the listener. Therefore, according to the voice characteristics of a leader, the cognition of the members may change, and, the degree to which the members are influenced and motivated will be different. This study identifies how members may perceive a speech differently according to the different voice characteristics of leaders in formal speeches. Further, different perceptions about voices will influence members' cognition of the leader, for example, in how trustworthy they appear. The study analyzed recorded speeches of leaders, and extracted features of their speaking style through digital speech signal analysis. Then, parameters were extracted and analyzed by the time domain, frequency domain, and spectrogram domain methods. We also analyzed the parameters for use in Natural Language Processing. We investigated which leader's voice characteristics had more influence on members or were more effective on them. A person's voice characteristics can be changed. Therefore, leaders who seek to influence members in formal speeches should have effective voice characteristics to motivate followers.

One major and time-consuming task in fish production is obtaining an accurate estimate of the number of fish produced. In most Nigerian farms, fish counting is performed manually. Digital image processing (DIP) is an inexpensive solution, but its accuracy is affected by noise, overlapping fish, and interfering objects. This study developed a catfish recognition and counting algorithm that introduces detection before counting and consists of six steps: image acquisition, pre-processing, segmentation, feature extraction, recognition, and counting. Images were acquired and pre-processed. The segmentation was performed by applying three methods: image binarization using Otsu thresholding, morphological operations using fill hole, dilation, and opening operations, and boundary segmentation using edge detection. The boundary features were extracted using a chain code algorithm and Fourier descriptors (CH-FD), which were used to train an artificial neural network (ANN) to perform the recognition. The new counting approach, based on the geometry of the fish, was applied to determine the number of fish and was found to be suitable for counting fish of any size and handling overlap. The accuracies of the segmentation algorithm, boundary pixel and Fourier descriptors (BD-FD), and the proposed CH-FD method were 90.34%, 96.6%, and 100% respectively. The proposed counting algorithm demonstrated 100% accuracy.

With the advances in Information Technology (IT), users can download or stream copyrighted works, such as videos, music, and webtoons, at their convenience. Thus, the frequency of use of copyrighted works has increased. Consequently, the number of unauthorized copies and sharing of copyrighted works has also increased. Monitoring is being conducted on sites suspected of conducting copyright infringement activities to reduce copyright holders' damage due to unauthorized sharing of copyrighted works. However, suspected copyright infringement sites respond by changing their domains or blocking access requests. Although research has been conducted for improving the effectiveness of suspected copyright infringement site detection by defining suspected copyright infringement sites' response techniques as a lifecycle step, there is a paucity of studies on automation techniques for lifecycle detection. This has reduced the accuracy of lifecycle step detection on suspected copyright infringement sites, which change domains and lifecycle steps in a short period of time. Thus, in this paper, an automated detection technique for suspected copyright infringement sites is proposed for efficient detection and response to suspected copyright infringement sites. Using our proposed technique, the response to each lifecycle step can be effectively conducted by automatically detecting the lifecycle step.

Cyberattacks penetrate the server and perform various malicious acts such as stealing confidential information, destroying systems, and exposing personal information. To achieve this, attackers perform various malicious actions by infecting endpoints and accessing the internal network. However, the current countermeasures are only anti-viruses that operate in a signature or pattern manner, allowing initial unknown attacks. Endpoint Detection and Response (EDR) technology is focused on providing visibility, and strong countermeasures are lacking. If you fail to respond to the initial attack, it is difficult to respond additionally because malicious behavior like Advanced Persistent Threat (APT) attack does not occur immediately, but occurs over a long period of time. In this paper, we propose a technique that detects an unknown attack using an event log without prior knowledge, although the initial response failed with anti-virus. The proposed technology uses a combination of AutoEncoder and 1D CNN (1-Dimention Convolutional Neural Network) based on semi-supervised learning. The experiment trained a dataset collected over a month in a real-world commercial endpoint environment, and tested the data collected over the next month. As a result of the experiment, 37 unknown attacks were detected in the event log collected for one month in the actual commercial endpoint environment, and 26 of them were verified as malicious through VirusTotal (VT). In the future, it is expected that the proposed model will be applied to EDR technology to form a secure endpoint environment and reduce time and labor costs to effectively detect unknown attacks.

A newfangled family of Nyquist-I pulses is proposed and named improved dual sinc pulse (IDSP). The IDSP is designed to improve performance in orthogonal frequency division multiplexing (OFDM)-based systems. The IDSP is a generalization of the dual sinc pulse (DSPP). This is because the DSP was formulated for α = 1 whereas the IDSPP is valid for 0 ≤ α ≤ 1. The behavior of the IDSP is promising in terms of its frequency and time domain responses. Theoretical and numerical outcomes indicate that the IDSP outperformed other existing pulses applied in OFDM-based systems for various key evaluation metrics.

In a device-to-device (D2D) underlaid cellular network, there exist two types of co-channel interference. One type is inter-layer interference caused by spectrum reuse between D2D transmitters and cellular users (CUEs). Another type is intra-layer interference caused by spectrum sharing among D2D pairs. To mitigate the inter-layer interference, we first derive the interference limited area (ILA) to protect the coverage probability of cellular users by modeling D2D users' location as a Poisson point process, where a D2D transmitter is allowed to reuse the spectrum of the CUE only if the D2D transmitter is outside the ILA of the CUE. To coordinate the intra-layer interference, the spectrum sharing criterion of D2D pairs is derived based on the (signal-to-interference ratio) SIR requirement of D2D communication. Based on this criterion, D2D pairs are allowed to share the spectrum when one D2D pair is far from another sufficiently. Furthermore, to maximize the energy efficiency of the system, a resource allocation scheme is proposed according to weighted graph coloring theory and the proposed ILA restriction. Simulation results show that our proposed scheme provides significant performance gains over the conventional scheme and the random allocation scheme.