Advances on P2P, Parallel, Grid, Cloud and Internet Computing

In this work, we present a distributed event-based awareness approach for P2P groupware systems. The awareness of collaboration is achieved by using primitive operations and services that are integrated into the P2P middleware. We propose an abstract model for achieving these requirements and we discuss how this model can support awareness of collaboration in mobile teams. In this paper, we present a Fuzzy Peer Coordination Quality System (FPCQS) for P2P networks according to four parameters. We consider Time for Finishing the Required Task (TFRT) as a new parameter. We evaluated the performance of proposed system by computer simulations. The simulation results show that that when GS is increased, the PCQ is increased. But, by increasing PCC, the PCQ is decreased. When the PM is 50 units, the PCQ is the best. Considering the effect of TFRT parameter, we found that when TFRT is increased, the PCQ is decreased.

Vehicular Ad hoc Networks (VANETs) have gained great attention due to the rapid development of mobile internet and Internet of Things (IoT) applications. With the evolution of technology, it is expected that VANETs will be massively deployed in upcoming vehicles. In addition, ambitious efforts are being done to incorporate Ambient Intelligence (AmI) technology in the vehicles, as it will be an important factor for VANET to accomplish one of its main goals, the road safety. In this paper, we propose an intelligent system for safe driving in VANETs using fuzzy logic. The proposed system considers in-car environment data such as the ambient temperature and noise, vehicle speed, and driver’s heart rate to assess the risk level. Then, it uses the smart box to inform the driver and to provide better assistance. We aim to realize a new system to support the driver for safe driving. We evaluated the performance of the proposed system by computer simulations and experiments. From the evaluation results, we conclude that the vehicle’s inside temperature, noise level, vehicle speed, and driver’s heart rate have different effects on the assessment of risk level.

The Wireless Mesh Networks (WMNs) are becoming an important networking infrastructure because they have many advantages such as low cost and increased high speed wireless Internet connectivity. In our previous work, we implemented a Particle Swarm Optimization (PSO) and Hill Climbing (HC) based hybrid simulation system, called WMN-PSOHC, and a simulation system based on Genetic Algorithm (GA), called WMN-GA, for solving node placement problem in WMNs. Then, we implemented a hybrid simulation system based on PSOHC and distributed GA (DGA), called WMN-PSOHC-DGA. In this paper, we analyze the performance of WMNs using WMN-PSOHC-DGA simulation system considering Constriction Method (CM) and Linearly Decreasing Vmax Method (LDVM). Simulation results show that a good performance is achieved for CM compared with the case of LDVM.

The growth in sensor networks and importance of active devices in the physical world has led to the development of Wireless Sensor and Actor Networks (WSANs). WSANs consist of a large number of sensors and also a smaller number of actors. Whenever there is any emergency situation i.e., fire, earthquake, flood or enemy attack in the area, sensor nodes have the responsibility to sense it and send information towards an actor node. According to these data gathered, the actor nodes take a prompt action. In this work, we consider the actor node selection problem and propose a fuzzy-based system (FBS) that based on data provided by sensors and actors selects an appropriate actor node to carry out a task. We use 4 input parameters: Degree of Centrality (DoC), Distance to Event (DE), Power Consumption (PC) and Number of Sensors per Actor (NSA) and the output parameter is Actor Selection Decision (ASD).

Due to the depletion of fossil fuel and global warming, the incorporation of alternative low carbon emission energy generation becomes crucial for energy systems. The wind power is a popular energy source because of its environmental and economic benefits. However, the uncertainty of wind power, makes its incorporation in energy systems really difficult. To mitigate the risk of demand-supply imbalance by wind power, an accurate estimation of wind power is essential. Recognizing this challenging task, an efficient deep learning based prediction model is proposed for wind power forecasting. In this proposed model, Wavelet Packet Transform (WPT) is used to decompose the wind power signals. Along with decomposed signals and lagged inputs, multiple exogenous inputs (calendar variable, Numerical Weather Prediction (NWP)) are used as input to forecast wind power. Efficient Deep Convolution Neural Network (EDCNN) is employed to forecast wind power. The proposed model’s performance is evaluated on real data of Maine wind farm ISO NE, USA.

Nowadays, Wireless Sensor Networks (WSNs) are facing various challenges. Cost efficiency, low energy consumption, reliable data communication between nodes and security are the major challenges in the field of WSNs. On the other hand, blockchain is also a very hot domain in this era. Blockchain has a remedy for some challenges, which are faced by the WSNs, e.g., secure data transactions and trustworthiness, etc. By keeping in mind the security issues, we induce blockchain into the WSNs. In short, we have proposed a trust model to avoid the malicious attacks and keep the transact data using the blockchain property of immutability. Moreover, an enhanced version of Proof of Stack (PoS), i.e., the Proof of Authority (PoA) consensus mechanism is being used to add a new node in the network. Additionally, the smart contract is also written to check the working status of nodes. Simulations are performed in order to record the transaction cost and execution cost.

The data sharing is the claim of actual scholars datasets to share and reuse in the future from any domain. The rise of blockchain technology has to increase universally and enhancement in share and reuse of scholars datasets. Despite there are numbers of security management frameworks for share data securely. However, those frameworks is a centralize based to make data share digitally. Its has restriction and owned by third party authority. The access and reuse of research datasets have a variety of issues it misinterpretation. In this aspect, the researcher or publisher has not to share data publicly due to reuse and perceive the risk in a data sharing environment. Preparing and storing data is difficult in contents sharing. To overcome the limitation and restriction, we proposed distributed data sharing management based on blockchain network (peer to peer P2P network). To signify on Ethereum framework, we proposed the case study of data sharing on the Ethereum smart contract platform to achieve the access.

The Internet of Things (IoT) industry is growing very fast to transform factories, homes and farms to make them automatic and efficient. In the past, IoT is applied in different resilient scenarios and applications. IoT faces a lot of challenges due to the lack of computational power, battery and storage resources. Fortunately, the rise of blockchain technology facilitates IoT devices in security solutions. Nowadays, blockchain is used to make reliable and efficient communication among IoT devices and emerging computing technologies. In this paper, a blockchain-based secure service provisioning scheme is proposed for Lightweight Clients (LCs). Furthermore, an incentive mechanism based on reputation is proposed. We used consortium blockchain with the Proof of Authority (PoA) consensus mechanism. Furthermore, we used Smart Contracts (SCs) to validate the services provided by the Service Providers (SPs) to the LCs, transfer cryptocurrency to the SPs and maintain the reputation of the SPs. Moreover, the keccak256 hashing algorithm is used for converting the data of arbitrary size to the hash of fixed size. The simulation results show that the LCs receive validated services from the SPs at an affordable cost. The results also depict that the participation rate of SPs is increased because of the incentive mechanism.

Wireless Sensor Network (WSN) is a network of nodes connected through a wireless channel. The sensor nodes in the network are resource constrained in terms of energy, storage and computational power. Node failure is a common phenomenon, which occurs due to environmental factors, adversary attacks, draining of battery power, etc. After node failure, recovery is challenging that needs a strong mechanism. In this paper, Blockchain-based Node Recovery (BNR) scheme for WSNs is proposed. In BNR scheme, recovery of failed nodes is on the basis of node degree. The working mechanism of the scheme is that firstly, the failed nodes are detected using the state (active or inactive) of Cluster Heads (CHs). In the second phase, the recovery process is initiated for inactive nodes. The main purpose of this step is to recover the failed CH, which ultimately results in restoring the active states of its member nodes. NodeRecovery Smart Contract (SC) is written for the purpose. The cost analysis for NodeRecovery is also performed in the proposed work. Moreover, the security analysis is performed to ensure the security of the proposed scheme. Effectiveness of the proposed model is shown by the simulation results.

Existing building automation solutions suffer from the drawback of causing high engineering, commissioning, and installation efforts. Furthermore, the existing approaches are error-prone and require a lot of manual interaction between the involved parties. This article presents the results of an analysis for moving towards autonomous plug&use functionality by exploiting external data sources merged and maintained in a central information model, e.g., device information obtained via the scanning of NFC or QR tags, functional planning information, network topology information, wiring plan information, etc. In this context, the autonomous plug&use functionality aims at fully automating the commissioning process of devices of a building automation system by deriving input parameters from external data sources thereby minimizing manual user efforts.

Assessing Mission-Critical systems is non-trivial, even more when Commercial-Off-The-Shelf (COTS) software tools, which have not been developed following custom reliability rules, are adopted. A satisfactory process of standard certification brings in a model estimation of system reliability. However, its evaluation requires in input reliability data of subsystem units, which are quite difficult to obtain. A practical issue, in fact, concerns a general lack of detailed statistical evaluations coming from on-field experiences. While, on the hardware side, the research community gave an effective contribution, on the software side, there is still work to do. An example is represented by the Cluster Resource Manager (CRM) software running on top of clustered systems, which is responsible of orchestrating fail-over operations. To the best of our knowledge, for such a component there isn’t any reliability evaluation based on field experiences.In this work, a particular CRM, namely Pacemaker, was tested to estimate the fail-over success probability in the occurrence of different type of resource outages. Pacemaker is one of the most accepted CRM and is used in several Critical Infrastructure (CI) contexts to ensure high availability of their Industrial Control System (ICS). Our experiments have been conducted on a real clustered ICS, the Train Management System (TMS) of Hitachi Ansaldo STS.

Given the enormous interest in self-driving cars, Vehicular Ad hoc NETworks (VANETs) are likely to be widely deployed in the near future. Cloud computing is also gaining widespread deployment. Marriage between cloud computing and VANETs, would help solve many of the needs of the drivers, law-enforcement agencies, traffic management, etc. In this paper, we propose a secure and distributed architecture for vehicular cloud which uses the capabilities of vehicles to provide various services such as parking management, accident alert, traffic updates, cooperative driving, etc. Our architecture ensures privacy of vehicles, and supports scalable and secure message dissemination using vehicular infrastructure.

Various different measures of textual similarity exist including string-based, corpus-based, knowledge-based, and hybrid-based measures. To our knowledge, none of them examine the textual connotation of two different sentences for the purpose of establishing whether they express a similar opinion. Connotation, within the context of this work, is the negative, positive, or neutral emotional meaning of words or phrases. In this paper we define a new type of a similarity measure mathematically, namely connotation similarity. It evaluates how similar the emotional meanings of two sentences are using opinion mining, which is also known as sentiment analysis. We compare two different algorithms of our definition of connotation similarity against various algorithms of cosine similarity using one dataset of 100 pairs of sentences, and another dataset of 8 pairs. We show how connotation similarity can be used on its own to indicate whether two sentences express a similar opinion, and also how it can be used to improve other similarity measures such as cosine similarity.

In order to make up the disadvantages of transfer learning methods, a cotton and flax fibred classification model with fusion of deep features transfer learning is proposed. First, the proposed model utilizes VGG16 and InceptionV3 to extract deep features from cotton and flax fiber images. Next, using spatial feature fusion algorithm, the model merges the deep features extracted from different networks. Finally, the generalized deep feature is used to train SoftMax classifier, thereby achieving accurate detection of cotton and flax fibers. In testing datasets which have 4008 images, the cotton and flax Fiber classification accuracy, sensitivity and specificity of the proposed model are 0.978, 0.969, 0.985 respectively. The experiments demonstrate that the proposed model outperforms the state-of-the-art model in the same hardware environment. The results show that the proposed model can detect cotton and flax Fiber with high accuracy.

Document automatic summarization technology is a method that refines documents and generates summaries representing the whole document to help people quickly extract important information. Aiming at solving lack of semantic information in document abstracts, this paper proposed a weighted hybrid document summary model based on LDA. This model obtains the theme distribution probability through analysing the document. Firstly, we used the FCNNM (Fine-grained Convolutional Neural Network Model) extract the semantic features, then search the surface information of the text from heuristic rules, including the length, location of the sentence and TF-IDF of the words in the sentence, and weighted to calculate the sentence score. Finally, used the greedy algorithm to select the sentence to form the abstract. Experiments show that the proposed model can effectively compensate for the lack of semantics between abstract sentences and text in traditional algorithms, effectively reduce the high redundancy in document abstracts and improve the quality of abstracts.

The function-as-a-service (FaaS) service model has been gaining adoption at a fast pace. In this service model, cloud applications are structured as self-contained code modules called functions that are instantiated on-demand, and billing is based on the number of function invocations and on function execution time. Developers are attracted to FaaS because it promises to remove two drawbacks of the traditional IaaS and PaaS service models, the need to provision and manage infrastructure, and the need to pay for unused resources. In practice, however, things are a little less rosy: developers still have to choose the amount of memory allocated to functions, and costs are less predictable, especially because they are tied to function performance. This work investigates performance and cost variations within and across FaaS providers. Our results show that performance and cost can be significantly affected by the choice of memory allocation, FaaS provider, and programming language: we observed differences of up to $$8.5\times $$ in performance and 67 $$\times $$ in cost between providers (with the same language and memory size), and $$16.8\times $$ in performance and $$67.2\times $$ in cost between programming languages (with the same provider and memory).

The task of streaming video files or sources in data centers is challenging since data streaming must be implemented at a constant rate without any interruption. The task of streaming requires proper storage space, network bandwidth and efficient algorithms that tackle the changing environmental conditions. In this paper, we conduct research on the optimality of incurring delay and consuming bandwidth when video streaming is established in data center network (DCN) architectures. We first observe the behavior of such traffic patterns in eight server rack and sixteen server rack architectures and compare the simulation results. We then expand our study to observe the behavior of higher complexity data center network. In our study, we attempted to include realistic situations such as appropriate Proxy servers, adaptive bit rate streaming (ABS) feature, and virtual private network (VPN) security feature. We also conduct a study on high-resolution video and multicast traffic in such environments.

With the ongoing rise and ease-of-use of services provided by major cloud providers, many enterprises migrate their infrastructure and applications to cloud platforms. However, the success of using many and diverse services leads to more attack vectors for potential attackers to exploit, that again leads to more difficult, complex and platform-specific security architectures. This paper is an attempt to remedy this problem by proposing simplified cloud security using machine learning techniques. This leads to a more general architecture that uses classifiers such as decision trees and neural networks, trained with data logged by cloud applications. More specifically, we collected easy-to-interpret access logs from multiple web applications which are often the only kind of security information available on various services on such platforms. The results show a more flexible approach, that was experimentally validated on cloud platforms and improve detection speed, using neural networks and J48 decision trees, up to 26–47 times while still maintaining an accuracy of 98.47% and 97.71% respectively.

The present work faces the problem of automatic classification and representation of unstructured texts into the Cultural Heritage domain. The research is carried out through a methodology based on the exploitation of machine-readable dictionaries of terminological simple words and multiword expressions. In the paper we will discuss the design and the population of a domain ontology, that enters into a complex interaction with the electronic dictionaries and a network of local grammars. A Max-Ent classifier, based on the ontology schema, aims to confer to each analyzed text an object identifier which is related to the semantic dimension of the text. Into this activity, the unstructured texts are processed through the use of the semantically annotated dictionaries in order to discover the underlying structure which facilitates the classification. The final purpose is the automatic attribution of POIds to texts on the base of the semantic features extracted into the texts through NLP strategies.

When it comes to address challenges in the area of distributed and parallel computing, Multi-Agent Systems (MAS) are emerging as a key architecture. Characterized as a collection of autonomous software (agents) which are able to cooperate in a distributed environment, MAS-based applications have proven capabilities when using cognitive processes, reasoning and knowledge representation in order to develop functionality related to complex and dynamic scenarios where the contribution of a single agent is computationally limited. In this paper, we propose an industrial platform which fully supports the development, deployment and maintenance cycle of MAS-based applications.

Social networks have renewed the ways audiences experience art and its spaces. The phenomenon concerns visitors, communicating their art experience through social media, and the artistic institutions, communicating their spaces and events. Sharing contents is a practice of fruition that allows the experience to be textualized. Our research focuses on how Igers represent themselves and their experience at museums, through a qualitative and quantitative description of the data collected. Our approach can support art institutions’ communication strategies.

This study follows a multidisciplinary approach which combines text mining techniques, data visualization, and educational research to focus online knowledge communities. Recently, social media shifted information from “official” source to user-generated content changing knowledge building practices. Starting from a large textual dataset, created using mining techniques, these authors build an interactive visual tool to explore conversation of an Italian Facebook group. The aim of this research is to build a web-based tool, which allows exploration and navigation through conversations and topics, to understand the significance of interactions in the research community. Results show that members’ participation in the community conversations has grown in the course of time. Comments published from 1 January 2012 to 31 December 2016 allow us to identify seven topics for different fields of interest in the ROARS community. Despite positive results, further investigations are required to give weight to the empirical evidence.

Transesophageal echocardiography and CT imaging have been used to provide definite diagnoses of cardiac diseases, such as angina and myocardial infarction. Transesophageal echocardiography has been performed by manually adjusting probe depth and ultrasound irradiation angle while referring to the echo image. However, it is difficult to grasp the three-dimensional (3D) position of the heart with echo images alone. Moreover, this takes a long time and puts a heavy burden on patients and doctors. Therefore, we propose a new method in order to smoothly create a preoperative plan. The proposed method replaces conventional transesophageal echocardiography with CT images. The proposed system can inspect CT images interactively and with a shorter examination time. Moreover, unlike transesophageal echocardiography, there is no burden on the patient when using the proposed method.

3D shape similarity seach have been studied for detecting or finding a specified 3D model among 3D CAD model database. We propose to use 3D point data for the search, because it has become easier to obtain 3D point data by photographs. CAD data is converted into 3D point data in advance. Then, using machine learning, we attempt to match those data with the 3D point data acquired in the field. It can be expected that the accuracy of matching is improved by directly handling 3D data. As a preliminary trial, we have tried to clasify 10 kinds of chair models in represented as 3D point data with a machine learning approach. It was suggested that 3D shape matching between 3D point data is possible by our proposed method as the result.

IT infrastructures have changed to cloud computing since this model allows omnipresent and on-demand access to share computing resources. Such resources can be quickly provisioned and released with minimal management effort. However, cloud computing has been challenged to be scalable as regards the infrastructure of the systems. These challenges are reinforced in distance-learning environments where cloud infrastructure is adopted, and the supporting off-campus units, located in precarious regions, suffer for poor infrastructure. Consequently, fog-cloud models have emerged to improve application performance using local prepossessing of data and analysis of communication requirements. The present study envisages the solution of a problem in a Brazilian university that offers distance education. A low cost technological solution is proposed using the Fog-Cloud computing paradigm. In order to evaluate the proposal, an experiment using real data was carried out and the results obtained point to viability of the proposal.

Nowadays, it has being verified some interesting improvements in I/O architectures. This is an essential point to complex and data intensive scalable applications. In the scientific and industrial fields, the storage component is a key element, because usually those applications employ a huge amount of data. Therefore, the performance of these applications commonly depends on some factors related to time spent in execution of the I/O operations. In this paper we present a research characterization on I/O improvements related to the storage targeting high-performance computing (HPC) and data-intensive scalable computing (DISC) applications. We also evaluated some of these improvements in order to justify their concerns with the I/O layer. Our experiments were performed in the Grid’5000, an interesting testbed distributed environment, suitable for better understanding challenges related to HPC and DISC applications. Results on synthetic I/O benchmarks, demonstrate how to improve the performance of the latency parameter for I/O operations.

Home air environments are affected by several pollutant sources. Among them, fine dust is the most dangerous pollutant, which can cause serious health problems. In general, residents in East-Asian countries trust fine dust information on weather forecast portals. However, the quality at air inside homes is very different from forecasted, as indoor air conditions cannot be observed easily. Although fine dust values can be observed in an air purifier, these readings can hardly be trusted, because one air purifier in a living room cannot determine the fine dust density values of other rooms or places. Therefore, this paper presents a home fine-dust monitoring system based on selected places applicable in any environments. By using the FRIBEE white Arduino board (built-in XBee shield Arduino board) and fine dust sensors, a star network topology is developed and each fine dust density value is determined by the fine dust sensors. Utilizing XBee and the Bluetooth network, the author examined fine dust density values on an Arduino serial monitor and Android applications in real time. To obtain a moderate result, two places in a home were selected to observe the fine dust density values. Moreover, indoor fine dust density values were compared with fine dust density values of weather forecast portals. The proposed systems can be used as a low-cost tool for home (or other indoor environments) fine dust monitoring.

In this work we present a system for the automatic analysis of text comments related to food products. Systematic analysis means allowing an analyst to have at a glance all the needed aggregated data and results that summarize the meaning of hundreds or thousands of comments, written in natural language. The analysis of the comments, and therefore the choices of the consumers, can therefore constitute a patrimony of very high value for the companies of the sector.At this aim we implemented a system, developed in Python. It uses the state of the art libraries of processing texts written in natural language, because the messages in natural language collected on the domain of food are written in Italian language.

This work presents a model for human activity recognition, through an IoT paradigm, using location and movement data, generated from an accelerometer. The activities of five individuals from different age groups were monitored, utilizing IoT devices, using the activities of four of these individuals to train the model and the activities of the remaining individual for test data. For the prediction of the activities, the Extra Trees algorithm was used, where the results of 81.16% accuracy were obtained when only movement data were used, 92.59% when using both movement and location data, and 97.56% when using movement data and synthetic location data.

The increase of the number and different types of devices within the Internet of Things context has brought several challenges over time. One of them is how to support heterogeneous device interoperability in IoT environments. This work proposes an architecture, called Omniconn, to tackle some of these communication issues. Utilizing the microservice approach, tests were performed with devices communicating through the Zigbee, Bluetooth LE, and Wi-fi protocols. The results were compared with tests using the same protocols in isolation. It was possible to perceive a low percentage of timeouts and invalid packages. In some cases, the use of multiple protocols had presented a lower performance compared with the isolated experiments, using Omniconn in a testbed environment was considered feasible. On the other hand, the majority of tests pointed out an enhancement in the average response time and number of requests, which reached 26% and 12% respectively.

In the IoT+Fog+Cloud architecture, with the ever increasing growth of IoT devices, allocation of IoT devices to the Fog service providers will be challenging and needs to be addressed properly in both the strategic and non-strategic settings. In this paper, we have addressed this allocation problem in strategic settings. The framework is developed under the consideration that the IoT devices (e.g. wearable devices) collecting data (e.g. health statistics) are deploying it to the Fog service providers for some meaningful processing free of cost. Truthful and Pareto optimal mechanisms are developed for this framework and are validated with some simulations.

Counterfeit medicines are increasing day by day and these medicines are damaging the health of people. Drug Regulatory Authorities (DRAs) are trying to overcome this issue. Synchronized electronic medicine record can mitigate this risk. We proposed a decentralized Blockchain (BC) based medicine licensing and authentication system to stop production of counterfeit medicines. Our proposed system provides a convenient way to register medicines by manufacturers with DRA. Vendors will also be registered with DRA, they are intermediate who buy from manufacturer and sale to customers. Furthermore, every transaction between manufacturer and vendor will be saved to BC. We used Proof of Collaboration (PoC) as a consensus mechanism, the manufacturer deals with more vendors will have more mining p ower. DRA has a different department, i.e., Licensing Department (LD), regulatory Department (RD), and Quality Control Department (QAD). These departments perform many actions, which will be saved in the BC database. LD registers manufacturers and vendors, RD imposes rules and QAD makes random checks to test the quality of medicines. We also propose manufacturer profile scheme, QAD rates manufacturer according to its quality of medicines from the feedback of users. Moreover, we provide an interface to the users, through which they can check the authenticity of medicine. We compare results of traditional licensing system with BC based licensing system.

With the phenomenon of code reuse is becoming more widespread in the same malicious family, this paper proposed a method to detect malicious code using a novel neural net. To implement our proposed detection method, malicious code was transformed into RGB images according to its binary sequence. Then, because of code reuse features can be revealed in the image, the images were identified and classified automatically using a flexible and lightweight neural net. In addition, we utilized dropout algorithm to address the data imbalance among different malware families. The experimental results demonstrated that our model performs well in accuracy and rate of convergence as compared with other models.

Preprocessing is a very important step in side channel analysis. The quality of the collected power traces seriously affects the efficiency of side channel analysis. Therefore, the preprocessing of Wavelet Transform (WT) and Wavelet Packet Denoising (WPD) are widely used. However, WT has certain defects in characterizing detail information of power traces. The threshold of WPD is not universal and adaptive. In order to solve these problems, it provides a preprocessing of power traces by combining WPD and Singular Spectrum Analysis (SSA), which takes advantage of the former to resolve the power consumption data, and the latter is used to extract the information of the low frequency and high frequency parts. Then, according to the fluctuation trend of singular entropy, the key information contained in the two parts is extracted adaptively, so as to improve the quality of power traces. Finally, through the selection of plaintext attack on the SM4 algorithm implemented by hardware, it can improve the efficiency of Correlation Power Analysis (CPA).

In the era of big data, annotated text data is a scarce resource. The annotated important semantic information can be used as keywords in text analysis, mining and intelligent retrieval, as well as valuable training and testing sets for machine learning. In the analysis, mining and intelligent retrieval of Traditional Chinese Medicine (TCM) patents, similar to Chinese herbal medicine name and medicine efficacy, disease name is also an important annotation object. Utilizing the characteristics of TCM patent texts and based on co-training method in machine learning, this paper proposes a method of annotating disease names from TCM patent texts. Experiments show that this method is feasible and effective. This method can also be extended to annotate other semantic information in TCM patents.

In the era of artificial intelligence and big data, a large number of legal case texts have been accumulated in the process of law enforcement of marine rights protection. These case texts contain a lot of important information, such as the time, place, person, event, judgment body, judgment result and so on. The annotation of these semantic information is an important link in text analysis, mining and retrieval of sea-related cases. In this paper, a semantic annotation method based on collaborative training for maritime legal texts is proposed. The experimental results show that the method is correct and feasible.

This paper presents a case study on data analytical platform deployment. The platform takes datasets continuously from social media platform, Facebook, and provide some insight information for sell personal in automotive industry in Thailand. The goal is to improve the sale performance of the sale team. Using information from the Facebook Graph API on the Facebook pages of an automotive industry dealer, the developed platform can then analyze such data and store the insight into the stored data system. When the users search for the related keywords or hashtags, the information in the form of graphs, images, and text can be returned. Both system performance experimental results, as well as the focus-group interviewing are to be presented to validate our proposed work.

In peer-to-peer networks for video live streaming, peers can share the forwarding load in two types of systems: unstructured and structured. In unstructured overlays, the graph structure is not well-defined, and a peer can obtain the stream from many sources. In structured overlays, the graph is organized as a tree rooted at the server and parent-child relationships are established between peers. Unstructured overlays ensure robustness and a higher degree of resilience compared to the structured ones. Indeed, they better manage the dynamics of peer participation or churn. Nodes can join and leave the system at any moment. However, they are less bandwidth efficient than structured overlays. In this work, we propose new simple distributed repair protocols for video live streaming structured systems. We show, through simulations and with real traces from Twitch, that structured systems can be very efficient and robust to failures, even for high churn and when peers have very heterogeneous upload bandwidth capabilities.

In recent years, Internet live video distribution has become popular and Internet distributors such as YouTubers have attracted attention. In Internet live distribution, the distributors often record and distribute themselves with a smartphone while moving, and deliver video via the internet. Regarding the trust between the distributor and the viewers, there is a social problem that the viewers are threaten or attacked by the distributor if there is no trust, and securing the trust in the Internet live distribution provides secure distributions to the distributor safe. It is possible to transmit information with trust. However, there has been no Internet live video distribution system considering the trust so far, and the distributor hides the face and distributes so that the surroundings and the surroundings are not displayed. We propose a trust-oriented Internet live distribution system for video processing, especially in this research, we designed rules for event-driven processing of the relationship between distributor information, trust environment, and video content in trust-oriented live video distributions.

Due to the recent popularization of machine learning, such a deep reinforcement learning as AlphaGO has advanced to analyze large-scale data and is attracting great attention. In deep reinforcement learning, users evaluate many functions in large-scale computer environments, including supercomputer and cloud systems. Cloud services can provide computer resources based on the scale of the computer environment desired by users. On the other hand, in conventional large-scale computer environment that only consists of CPUs or GPUs, the processing time greatly increases according to the scale of the calculation processing. In this paper, we propose a high-performance computing environment for deep reinforcement learning that links supercomputer and cloud systems. Our proposed system can construct a high-performance computing environment based on the scale of the computing process by the cooperation of the supercomputing and cloud systems with short physical distance and short network distance. In our evaluation of deep reinforcement learning using our proposed system, we confirmed that computer resources can be effectively used by allocating suitable processing for the supercomputer and the cloud according to the usage situations of the CPU, the GPU, and the memory.

We define continuous sensor data with difference cycles as “sensor data streams” and have proposed methods to collect distributed sensor data streams. However, our previous paper provides the simulation results only when the distribution of collection cycles is uniform. Therefore, This paper provides the additional simulation results in different distributions of collection cycles. Our additional simulation results show that our proposed method can equalize the loads of nodes even if the distribution of collection cycles are not uniform.

Due to the recent development of wireless broadcasting technologies, video data distribution systems in hybrid broadcasting environments has attracted great attention. In some video distribution methods for hybrid broadcasting environments, the server broadcasts some data pieces via the broadcasting channel and delivers them to all the clients to effectively reduce the interruption times. They fix the bandwidth allocation to predict when the clients finish receiving pieces. However, the server can change the bandwidth allocation flexibly in recent environments such as 5G. In this paper, we propose a mathematical model for video data distribution in flexible bandwidth hybrid broadcasting environments.

In this research, we investigated the influence on Kansei when placing Japanese traditional crafts in virtual reality space of Japanese-style room, Western-style room and Chinese-style room. We placed Japanese traditional crafts such as Fusuma (sliding door), Shoji (paper sliding door) and Tsuitate (screen) in each room style. We conducted a questionnaire survey on 14 subjects using 10 Kansei word pairs on the impression of Japanese traditional crafts placed in each room style.

Semantic similarity calculation between words is an important step of text analysis, mining and intelligent retrieval. It can help to achieve intelligent retrieval at the semantic level and improve the accuracy and recall rate of retrieval. Because of the particularity of TCM (Traditional Chinese Medicine) patents and the insufficiency of research, most of the current mainstream TCM patent retrieval systems are keywords-based, and the retrieval results are not satisfactory. In order to improve the intelligence level of TCM patent retrieval, to promote TCM innovation and avoid repetitive research, based on real TCM patent corpus, this paper utilizes the excellent feature learning ability of deep learning to build a neural network model, and gives a method to calculate the semantic similarity between words in TCM patents. The experimental results show that the proposed method is effective. In addition, this method can be extended to semantic similarity calculation in other domains.

With the continuous construction and development of China’s marine industry, the protection of marine rights has become a crucial issue. Usually, the knowledge of maritime law is oriented to legal practitioners, and the general public does not know much about it. Most of the persons responsible for maritime cases are mainly due to their weak legal consciousness and lack of understanding of marine laws. Therefore, it is very important to develop a knowledge assistant software which can popularize marine law to reduce the probability of illegal cases. With the help of abundant maritime judgment cases and relevant professional knowledge on the Internet, this paper will design and develop a maritime legal knowledge assistant software built on Android platform. It can not only provide timely, convenient, accurate and abundant information services for the vast number of maritime practitioners and functional departments such as maritime courts through advanced mobile service mode, it can also promote the reform of maritime judgment system.

The traditional collaborative filtering method based on matrix factorization regards the users preference as the inner product of users and items implicit features and has limited learning ability. Many studies have focused on the use of deep neural networks to mine the interaction relationship between implicit features, but the learning cost of deep neural networks is too large, and the model lacks interpretability. Therefore, a matrix factorization recommendation method based on multi-grained cascade forest is proposed. Replacing inner product by multi-grained cascade forest with deep structures, rather than deep neural networks, and explore the interactive relationship between users and items. The method experiments on real-world data sets and performs well when comparing with the state of the art methods.

Discussion forums are among the most common interaction tools offered by MOOCs. Nevertheless, due to the high number of students enrolled and the relatively small number of tutors, it is virtually impossible for instructors to effectively monitor and moderate them. For this reason, teacher-guided instructional scaffolding activities may be very limited, even impossible in such environments. On the other hand, students who seek to clarify concepts may not get the attention they need, and lack of responsiveness often favors abandonment. In order to mitigate these issues, we propose in this work a multi-attribute text categorization tool able to automatically detect useful information from MOOC forum posts including intents, topics covered, sentiment polarity, level of confusion and urgency. Extracted information may be used directly by instructors for moderating and planning their interventions as well as input for conversational software agents able to engage learners in guided, constructive discussions through natural language. The results of an experiment aimed at evaluating the performance of the proposed approach on an existing dataset are also presented, as well as the description of an application scenario that exploits the extracted information within a conversation agents’ framework.

A problem that students find in some subjects is to find online educational resources on their content that have adequate quality. This paper presents a tool that allows a teacher to create in a simple way educational resources using their own material or automatically extracted from repositories of open data or linked data. In addition, the tool offers a user management system that allows a student to register with a specific teacher, so that they can access all the resources published by a specific teacher. Educational resources can be downloaded from the tool in order to be used non-online. In the article, the proposal has been illustrated in the context of the subject of Software Engineering.

In the last decade, cloud development has grown exponentially and increasingly companies, administrative and educational institutions decide to make the leap and turn their solutions to a cloud platform. In particular, cloud technologies are applicable to educational contexts, especially in online learning, as online universities historically have their educational services installed on-premises, but the trend is to move them to the cloud. This is mainly because of the evident non-functional benefits of using cloud-based solutions, such as high availability, scalability and real-time responsiveness, which are impossible or too much costly to be achieved on-premises Indeed, these benefits can effectively support the current broad and demanding educational services, such as document sharing, communication, assessment, administrative procedures and reporting tools, which must be available to an increasing number of students and university staff anytime and anywhere. However, from the architectural point of view, cloud-based systems pose some challenges not existing in traditional on-premises systems, such as modularization and separation of services, which require additional work to guarantee performance and data protection against potential risks. In this paper, we focus on the assessment services provided by an innovative cloud-based educational system named TeSLA, which may reduce considerably the university costs and infrastructure maintenance, while offering flexible and effective e-assessment solutions. The ultimate goal of the paper is to propose a methodology to evaluate the TeSLA system from the underlying cloud infrastructure in terms of non-functional requirements.

Today, lifelong learning is fully integrated into our society. From the student point of view, lifelong learning has several characteristics that differentiate it from regular learning: domains of interest may be very broad; learning occurs in different depths; topics to study may be related both to work, family and leisure; students’ continuity cannot be guaranteed since their availability can be intermittent and little constant; a great dynamism is required in order to allow studying any topic, in any order, in the moment that best suit each student and at the best pace for everyone. Over 25 years ago some authors already claimed for moving towards innovative learning models, more personalized and where the students would take a more active role and would decide what to learn, when to learn and how to learn. Technology was not ready then to support this change of pedagogical paradigm, but it seems to be ready now. Thus, the technological context is set for facilitating a change of paradigm to promote lifelong learning. However, lifelong learners continue suffering from a model not adapted to their needs and preferences. This position paper discusses on the actual situation of lifelong learning from a critical point of view, analyzing some of the relevant literature and justifying the need to create new models that promote self-determination of students in the context of lifelong learning.

In this paper we introduce new mechanisms of dispute resolution as a helping tool in legal procedures for lawyers, mediators and judges with the objective to reach an agreement between the parties; in some situations. The primary objectives are the following: to apply algorithmic mechanisms to the solution of certain national and cross-border civil matters, including matrimonial regimes, successions and trusts, commercial law and consumer law, facilitating the agreement process among the parties; to demonstrate the efficacy of an algorithmic approach and apply it into the allocation of goods, or the resolution of issues, in disputes, leading the parties to a friendly solution before or during the trial comparison. We will focus on the algorithms used for resolving disputes that involve a division of goods between agents, e.g. inheritance, divorces and company law.

This paper proposes a hardware solution for the SHA-256 hash function offering a number of configurable architecture-level features. This flexibility allows for exploring various trade-offs between performance, area occupation, and power consumption. As confirmed by the experimental results, the approach succeeds in exposing the effects of different architectural configurations on the resulting implementation.

Crowd Sensing Network (CSN) uses sensor embedded mobile phones for the collection of data for some specific task which can effectively save cost and time. The quality of collected data depends on the participation level from all entities of CSN, i.e., service provider, service consumers and data collectors. In comparison with the centralized traditional incentive mechanisms devised for CSN, we have proposed a decentralized system model where incentives are used to stimulate the involvement among data collectors and motivate the participants to join the network. Moreover, the issue of privacy leakage is tackled by using AES128 technique. Furthermore, the system is evaluated through analyzing the gas consumption of all the smart contracts, whereas, the encryption technique is validated through comparing the execution time with base paper methods.

With the terms “Cloud Computing” we mean a model in which computation, software and storage resources do not require knowledge of the physical location and configuration of the system that provides the service in question by the end user. Thanks to the widespread use of the Internet, applications can now be used as services on the network. Many interesting areas of application are enjoying of these benefits, such as legal domain: the use of systems for conflict resolution can help lawyers, mediators and judges with the objective to reach an agreement between the parties. In this paper we introduce the architecture of a cloud-oriented web application, developed within the European project CREA (Conflict Resolution with Equitative Algorithms, Figure). Its objective is the application of algorithms to solve civil matters, into the allocation of goods, leading the parties to a friendly solution before or during the trial comparison.

In this paper we present potential advantages of an algorithmic approach in legal dispute resolution, introducing the CREA project with its objectives and methods. We describe a model that represents legal disputes by constraints satisfaction problems and two algorithmic strategies to solve them (Nash allocation and Egalitarian Equitative Allocation), focusing on inheritance disputes category. This approach is radically innovative because it helps lawyers and judges to set the legal procedure not as a parties’ dispute but as a process aiming to consensual agreement. We also report the experimental results to prove the effectiveness of our approach, considering a concrete legal case dispute.

In academic institutions, there is frequently the need to provide new services, in a cloud model, to be used either in teaching or research activities. One of the main decisions to be addressed is related to the cloud model to adopt (private, public or hybrid), and what the mixing of functionalities to use for the hybrid one. In this paper a new approach is proposed to define a methodology to serve as a tool to be used as decision support for the ICT manager in order to help him in this decision. A simple and intuitive graph representation has been used. The methodology has been tested with a real case study on the provisioning of a new e-learning service for the university of the authors. The work has to be considered a tentative approach as a starting point to further researches in the same direction.

The purpose of the study is to define and develop a methodology in order to evaluate the different levels of cloud Tourism implementation within Local Tourist Districts in Italy. The methodology evaluates whether and how the region is meeting the challenge posed by new technologies and to possibly identify performance value enhancing factors. The phenomenon observation is structured using a questionnaire referred to organizational typologies which should be submitted to all top level managers of all the analyzed districts. The acquired data should be processed using the factor analysis approach in order to identify the district organizational typologies.

One of the main advantages of cloud computing is elasticity, which allows to rapidly expand or reduce the amount of leased resources in order to adapt to load variations, guaranteeing the desired quality of service. Auto-scaling is an extensively studied topic. Making optimal scaling choices is of paramount importance and can help reduce leasing costs, as well as power consumption. This paper analyzes the current status of auto-scaling in the cloud ecosystem, considering recent literature contributions as well as existing auto-scaling solutions. Then it discusses possible research directions in this field, fostering the development of a methodology that, on the basis of suitably-defined performance parameters, can produce an optimal auto-scaling policy to be implemented using existing auto-scaling services and tools.

Cloud applications are by nature dynamic and must react to variations in use, and evolve to adopt new Cloud services, and exploit new capabilities offered by Edge and Fog devices, or within data centers offering Graphics Processing Units (GPUs) or dedicated processors for Artificial Intelligence (AI). Our proposal is to alleviate this complexity by using patterns at all stages of the Cloud application life-cycle: deployment, automatic service discovery, monitoring, and adaptive application evolution. The main idea of this paper is that it is possible to reduce the complexity of composing, deploying, and evolving Cross-Cloud applications using dynamic patterns.

The motivation for this transition comes from the fact that constantly maintaining a monolithic architecture has resulted in difficulties in keeping up in pace with new development approaches such as DevOps, calling for deployment several times a day. In contrast, microservices offer a more flexible option, where individual services comply with the single responsibility principle (SRP), and they can therefore be scaled and deployed independently. We propose a methodology, starting from a source code application, that provides a new architecture for the application oriented to microservices and deployable on Cloud.

Cloud technologies provide capabilities that can guarantee to the end user high availability, performance and scalability.However, the growing use of IoT technologies and devices, have made the applications not only more computationally intensive, but also data intensive. Because of this, dynamically scaling applications running on clouds can lead to varied and unpredictable results due to highly time-varying workloads distinguishes this new kind of applications. These applications are also often composed of different independent modules that could be easily moved across devices. Automatic scheduling and allocation of these modules is not an easy task, because there could be many conditions that prevent the design of a smart solutions. Thus determining appropriate scaling policies in a dynamic non-stationary environment is non-trivial, as a problem arises concerning resource allocation. Decision making about which resources should be added and removed, when the underlying performance of the resource is in a constant state of flux, becomes an issues. In this work we model both the applications and the infrastructure in order to formulate e Reinforcement Learning problem for automatically find the best configuration for the applications modules, taking into account the environment in which they are placed and the applications already running.

The DUST framework is a middleware platform developed to create software components that are distributable across heterogeneous networks. Developing and maintaining the component-based system and its configuration can become a challenge. In this paper, we present a web-based tool that allows engineers to visually model a DUST application graph in a CAD canvas. This graph is used by the back-end template generators to create the necessary configuration files and code for each module to implement in order to focus on the business logic. The current version of the tool is able to fully generate message objects and builder classes that are used for communication over DUST. Therefore, allowing the engineers to manage and update these objects through the GUI interface instead of error-prone coding.

As the Internet of Things (IoT) paradigm becomes omnipresent, so does fog computing, a paradigm aimed at bringing applications closer to the end devices, aiding in lowering stress over the network and improving latency. However, to efficiently place application tasks in the fog, task placement coordination is needed. In this paper, task placement in the fog and corresponding problems are addressed. We look at the fundamental issue of solving Multi-Objective Optimization problems and treat different techniques for distributed coordination. We review how this research can be used in a smart vehicle environment, and finish with some preliminary tests results.

Recent advances in the field of Neural Architecture Search (NAS) have made it possible to develop state-of-the-art deep learning systems without requiring extensive human expertise and hyperparameter tuning. In most previous research, little concern was given to the resources required to run the generated systems. In this paper, we present an improvement on a recent NAS method, Efficient Neural Architecture Search (ENAS). We adapt ENAS to not only take into account the network’s performance, but also various constraints that would allow these networks to be ported to embedded devices. Our results show ENAS’ ability to comply with these added constraints. In order to show the efficacy of our system, we demonstrate it by designing a Recurrent Neural Network (RNN) that predicts words as they are spoken, and meets the constraints set out for operation on an embedded device.

Over the last few years there has been a considerable amount of progress in the field of machine learning. Neural networks are common in academic literature and they are often used in engineering applications. A certain class of Artificial Neural Networks (ANN) is called Spiking Neural Networks (SNN), these are neural networks that utilise the specific time when a neuron fires, to calculate its output. Feedback loops with these kind of SNNs allow the execution of complex tasks in a compact manner. Due to their parallel character they are unquestionably suitable to be implemented on Field Programmable Gate Arrays (FPGA). The aim of this paper is to take a step towards creating a collision avoidance robot, which uses a SNN on an FPGA and Reinforcement Learning (RL) on an external device.

Many software-implemented control flow error detection techniques have been proposed over the years. In an effort to reduce their overhead, recent research has focused on selective approaches. However, correctly applying these approaches can be difficult. This paper aims to address this concern and proposes a new approach. Our new approach is easier to implement and is applicable on any existing control flow error detection technique. To prove its validity, we apply our new approach to the Random Additive Control Flow Error Detection technique and perform fault injection experiments. The results show that the selective implementation has approximately the same error detection ratio with a decrease in execution time overhead.

For autonomous navigation and robotic applications, sensing the environment correctly is crucial. Many sensing modalities for this purpose exist. In recent years, one such modality that is being used is in-air imaging sonar. It is ideal in complex environments with rough conditions such as dust or fog. However, like with most sensing modalities, to sense the full environment around the mobile platform, multiple such sensors are needed to capture the full 360-degree range. Currently the processing algorithms used to create this data are insufficient to do so for multiple sensors at a reasonably fast update rate. Furthermore, a flexible and robust framework is needed to easily implement multiple imaging sonar sensors into any setup and serve multiple application types for the data. In this paper we present a sensor network framework designed for this novel sensing modality. Furthermore, an implementation of the processing algorithm on a Graphics Processing Unit is proposed to potentially decrease the computing time to allow for real-time processing of one or more imaging sonar sensors at a sufficiently high update rate.

In smart cities, a myriad of devices is connected via Low Power Wide Area Networks (LPWAN) such as LoRaWAN. There is a growing need for location information about these devices, especially to facilitate managing and retrieving them. Since most devices are battery-powered, we investigate energy-efficient solutions such a Received Signal Strength (RSS)-based fingerprinting localization. For this research, we use a publicly available dataset of 130 426 LoRaWAN fingerprint messages. We evaluate ten different Machine Learning algorithms in terms of location accuracy, $$R^2$$ score, and evaluation time. By changing the representation of the RSS data in the most optimal way, we achieve a mean location estimation error of 340 m when using the Random Forest regression method. Although the k Nearest Neighbor (kNN) method leads to a similar location accuracy, the computational performance decreases compared to the Random Forest regressor.

Recent research focuses on how agents can learn to communicate with each other. This communication between the agents allows them to share information and coordinate their behaviour. Recent efforts have proven successful in these cooperative problems. A major problem we face in multi-agent reinforcement learning is the lazy agent problem, where some agents take advantage of the successful actions of other agents. This results in agents not being able to learn a functional policy. In this paper we will combine state-of-the-art methods to design an architecture to address cooperative problems using communication while also eliminating the lazy agent problem. We propose two approaches for learning to communicate that use value decomposition to address the lazy agent problem. We find that the additive version of value decomposition gives us results which exceeds the results of the state of the art.

Estimations indicate that up to a third of the power consumption of compressed air leaks is lost due to undetected leaks. The current methods of detection and localization of these leaks require intensive manual labor, making use of handheld devices. In addition, the added energy cost caused by these air leaks is concealed in the total cost of energy. These factors can explain why there is a limited commitment to detect and repair air leaks. To address this issue, we propose a solution which does not require manual labor in the process of detecting and locating pressurized air leaks. By equipping existing factory vehicles with a multi-modal sensing device containing a LIDAR, an ultrasonic microphone array and a camera, we are able to locate leaks in large industrial environments with high precision in 3D. With this proposed solution we aim to encourage the industry to proactively search for pressurized air leaks and therefore reduce energy losses costing a fraction of currently employed methods.

Location-based services are an essential aspect of many Internet of Things (IoT) applications. Due to low power requirements that these applications impose on their end devices, classic GNSS solutions are replaced by wireless localization via Low Power Wide Area Networks (LPWAN), e.g. Time Difference of Arrival (TDoA) localization with LoRaWAN. Usually, at least four gateways are required to obtain a reliable location estimate with TDoA. In this research, we propose to combine TDoA with Angle of Arrival (AoA) in a probabilistic way, using only two gateways. Our simulations demonstrate a 548 m mean error using TDoA between two gateways. Moreover, we reduce this mean error to 399 m when a single AoA estimate is added to the TDoA estimate.

Most of the vehicular industry applications require precise, reliable, and secure positioning with cm level accuracy. IEEE 802.11p and LTE-V exploit technological solutions to achieve communication vehicle to vehicle V2V, vehicles to infrastructure V2I, vehicle to pedestrians V2P, or vehicle to everything V2X. This paper presents the achievable localization accuracy of IEEE 802.11p and LTE-V. Cramer Rao Lower Bound for time difference of arrival localization for two different vehicular network scenarios is computed to determine bounds of accuracies of the technologies. Measurements are simulated assuming additive white Gaussian noise channel with a variance that depends on the geometry of the vehicular network. The simulation results show that IEEE 802.11p outperforms LTE-V for the conditions considered in this work. In addition to this, having network sites at both sides of the highway and considering the geometry between vehicles and network sites improves vehicle localization accuracy.

Modern cars contain numerous sensors that provide useful data in many different situations, but the interpretation of that data is cumbersome due to the different implementations of the Controller Area Network (CAN) messaging system. Hence, reverse engineering is needed in order to give sense to the internal sensor data of the car. Currently, reverse engineering of CAN data is an ongoing topic in research, but no method has been proposed yet to perform online reverse engineering. Therefore, this paper presents two methodologies. The first one elaborates on the online analysis of continuous signals, while the second one focuses on the reverse engineering of user-based signals, such as direction indicators and light switches. The results show that more research is needed in thoroughly benchmarking those methods with the current State of the Art. However, as the results are promising, this paper paves a way to a more scalable solution for reverse engineering in future applications.

Low-Power Wide Area Networks (LPWAN) for resilient Internet of Things (IoT) ecosystems come with unprecedented cost for the minimal load of communication. Long Range (LoRa) Wide Area Network (LoRaWAN) is a LPWAN which has a long range, low bit rate and acts as a connectivity enabler. However, making an efficient collaborative service of clock synchronization is challenging. In this paper we tackle two problems of effective robustness in LoRa network. First, current research typically focuses on the benefits of LoRa but ignores the requirement of reliability, which may invalidate the expected benefits. To tackle this problem, we introduce a novel time synchronization scheme for radically reducing usage of existing Aloha type protocol that handles energy consumption and service quality. Second, we look into the security space of LoRa network, i.e. channel selection scheme for the given spectrum. Attacks like selective jamming are possible in LoRa network because the entire spectrum space is not used, and utilization of few channels are comparatively higher. To tackle this problem, we present a channel hopping scheme that integrates cryptographic channel selection with the time notion for the current communication. We evaluate time synchronization and the channel hopping scheme for a real-world deployed peer to peer (P2P) model using commodity hardware. This paper concludes by suggesting the strategic research possibilities on top of this platform.

Perception of the world around is key for autonomous driving applications. To allow better perception in many different scenarios vehicles can rely on camera and LiDAR sensors. Both LiDAR and camera provide different information about the world. However, they provide information about the same features. In this research two feature based fusion methods are proposed to combine camera and LiDAR information to improve what we know about the world around, and increase our confidence in what we detect. The two methods work by proposing a region of interest (ROI) and inferring the properties of the object in that ROI. The output of the system contains fused sensor data alongside extra inferred properties of the objects based on the fused sensor data.

This paper describes a system to infer the brand of apple considering physical features of its flesh. The system comprises a hardware to examine the apple’s physical features and a software with convolutional neural network (CNN) to classify an apple into any brand. When a sharp metal blade cuts the piece of the apple flesh, the load and the sound are measured. From these data, the computer generates an image consisting of the sound spectrogram and the color bar expressing the load change. The sound spectrogram has rich features of the apple flesh. The image is inputted to CNN to infer the brand of apple. In the experiment part, the authors validated the proposed system. The goal of our study is to construct a system to estimate the texture such as crunchiness or crispness. The system is applicable to the quality management of the brand of apples. For example, one apple randomly chosen from many apples could be examined by the present system in order to check the texture quality of the flesh.

This paper describes a fundamental system to evaluate the welding performed by beginners. The authors took several pictures of metal plates welded by beginners, and then made image data. The image data is a part of welding joint in the picture. The authors extracted the welding partial image from the picture by hand. The extracted image data are divided into two categories. The one is “good” welding image and the other is “bad” one. The image was inputted to CNN to classify the images to “good” or “bad”. In the experiment, the validation of CNN was carried out. In the conclusion part, the result of the experiment and future works are discussed.

In this paper, we present two soft computing techniques, which are support vector regression (SVR) and fuzzy logic, to build an early warning model for detecting environmental pollution of waste-water in industrial zones. To determine the proper number of inputs for the model, we use an algorithm to find the embedding dimension space for a time series. Our proposed model, which has a high accuracy and short training time, to helps waste-water processing station operators take early action and avoid environmental pollution.

Correntropy is a dependence measure that goes beyond Gaussian environments and optimizations based on Minimum Squared Error (MSE). Its ability to induce a metric that is fully modulated by a single parameter makes it an attractive tool for adaptive signal processing. We propose a sparse modeling framework based on the dictionary learning technique known as K–SVD where Correntropy replaces MSE in the sparse coding and dictionary update subroutines. The former yields a robust variant of Orthogonal Matching Pursuit while the latter exploits robust Singular Value Decompositions. The result is Correntropy–based dictionary learning. The data–driven nature of the approach combines two appealing features in unsupervised learning—robustness and sparseness—without adding hyperparameters to the framework. Robust recovery of bases in synthetic data and image denoising under impulsive noise confirm the advantages of the proposed techniques.

Analyzing EEG signals can help us make implications about the user’s activities or even thoughts which can result in a myriad of applications. However, clinical EEG monitoring tools are expensive, often immobile and in need of professional supervision. Lately a couple of companies started the production of relatively cheap, easy-to-use, and mobile devices with significantly lower accuracy. In this paper, we intend to investigate the usability of these devices in recognizing selected basic activities e.g., winking, raising a hand etc., showing preliminary results on how clustering can prove to be an efficient method in labeling a low-quality EEG data set so that it could be used in supervised learning scenarios.

Development of Internet of Things (IoT) network brings new concept of Internet. The dramatic growth of IoT increased its usage. IoT network facilitates in several manners, more specifically, in access control and data sharing among IoT devices. However, it has many challenges, such as: security risks, data protection and privacy, single point of failure through centralization, trust and data integrity issues, etc. This work presented a blockchain based access control and sharing system. The main aim of this work is to overcome the issues in access control and sharing system in IoT network and to achieve authentication and trustworthiness. Blockchain technology is integrated with IoT, which simplifies the access control and sharing. Multiple smart contracts: Access Control Contract (ACC), Register Contract (RC), Judge Contract (JC), are used that provide efficient access management. Furthermore, misbehaviour judging method utilizes with penalty mechanism. Additionally, permission levels are set for sharing resources between users. Simulation results show the cost consumption. Bar graphs illustrate the transaction and execution cost of smart contracts and functions of main contract.

Nowaday’s energy demand and energy production are increasing. Renewable energy resources will play an important role in managing future production of electricity due to an increase in the development of societies. The centralized energy trading system faces a challenge in terms of fair energy distribution. Centralized existing energy trading system totally relies on a central system or third party, because the third party has many drawbacks in the form of record tampering or record altering. The fair transaction is the main issue in the energy trading sector. When the bitcoin is introduced in the market, the trust of Blockchain technology is increased. We proposed a Blockchain based energy trading system in peer to peer networks. Blockchain technology provides trust, security, and transparency for energy trading. In Blockchain technology, there is no necessary need of the third party in the energy supply sector. In our proposed paper, we facilitate the prosumer who produces renewable energy and sells surplus energy to the consumer. We achieved transparency, accuracy, efficiency in our proposed paper. Using a double auction process, we obtain low energy price and acheived consumer trust in energy trading.

With the acceleration of education informatization, the social demand for online examination papers is increasing. However, there are some problems in the generation of online examination papers. Firstly, it is impossible to randomly generate examination papers quickly. Besides, it is impossible to dynamically adjust examination papers according to test results. Thirdly, it is impossible to generate examination papers based on individual characteristics of students. In order to solve these problems, this paper proposes a new auto-generation examination paper model based on genetic algorithm. The model dynamically adjusts the difficulty factor of individual test questions by analyzing the online learning data and historical user test result data, and then guarantees the difficulty of generating examination papers in line with the changes in the current educational environment. The simulation results show that the algorithm improves the efficiency and accuracy of the generation examination paper, and effectively controls the difficulty coefficient of the examination paper.

In recent years, experts have considered the job of data scientists as the sexiest of 21st century. However, people skilled with data scientist’s expertise seem to be rare. This probably happens for the complex set of competences that this profession requires. In this paper, we deal with companies that are searching for data scientists to expand their workforce. Scraping data from the business-networking website LinkedIn, as for companies, we collected dimensions, sectors, kinds of employment, contract forms, working functions, and required skills. Our findings suggest that data scientist profession extends to several sectors but it is not yet consolidated. This condition intensifies the misconception about the skills required. Based on all this, we think that the role of higher institutions becomes fundamental, on the one hand to define data science as a discipline, and on the other to train young people for acquiring the set of skills needed.

It is very important to quickly solve system failures in a system operation. Some studies have proposed fault tolerance systems such as a flexible system architecture for dealing with system failures and automatic failure detection system. However, human identifies a system failure in many cases, and a support system to reduce the cost of trial and error for solving system failures is required. In this study, we propose an architecture for system recovery based on solution records on different servers. In the experiment using prototype, we confirm the feasibility of the proposed system.

Nowadays, online news platforms have become the main sources of news for many users. Hence, an urgent need arises to find a way to classify this news automatically and measure its validity to avoid spreading fake news. In this paper, we tried to simulate how humans, in real life, are dealing with news documents. We introduced a new way in which we can deal with the whole textual content of the news documents by extracting a number of characteristics of those texts and extracting a complex set of other metadata related features without segmenting the news documents into parts (title, content, date, source, etc.). Performances of nine machine learning algorithms in terms of Accuracies, Precision, Recall and F1-score are compared when using three different datasets obtaining much better result than the results in [1] and [2].