Vol. 41 · No. 5

Cloud computing provides hosted services (i.e., servers, storage, bandwidth, and software) over the internet. The key benefits of cloud computing are scalability, efficiency, and cost reduction. The key challenge in cloud computing is the even distribution of workload across numerous heterogeneous servers. Several Cloud scheduling and load-balancing techniques have been proposed in the literature. These techniques include heuristic-based, meta-heuristics-based, and hybrid algorithms. However, most of the current cloud scheduling and load balancing schemes are not content-aware (i.e., they are not considering the content-type of user tasks). The literature studies show that the content type of tasks can significantly improve the balanced distribution of workload. In this paper, a novel hybrid approach named Particle Swarm Optimization based Content-Aware Load Balancing Algorithm (PSO-CALBA) is proposed. PSO-CALBA scheduling scheme combines machine learning and meta-heuristic algorithm that performs classification utilizing file content type. The SVM classifier is used to classify users' tasks into different content types like video, audio, image, and text. Particle Swarm Optimization (PSO) based meta-heuristic algorithm is used to map user's tasks on Cloud. The proposed approach has been implemented and evaluated using a renowned Cloudsim simulation kit and compared with ACOFTF and DFTF. The proposed study shows significant improvement in terms of makespan, degree of imbalance (DI).

Cloud technology usage in nowadays companies constantly grows every year. Moreover, the COVID-19 situation caused even a higher acceleration of cloud adoption. A higher portion of deployed cloud services, however, means also a higher number of exploitable attack vectors. For that reason, risk assessment of the cloud environment plays a significant role for the companies. The target of this paper is to present a risk assessment method specialized in the cloud environment that supports companies with the identification and assessments of the cloud risks. The method itself is based on ISO/IEC 27005 standard and addresses a list of predefined cloud risks. Besides, the paper also presents the risk score calculation definition. The risk assessment method is then applied to an accounting company in a form of a case study. As a result, 24 risks are identified and assessed within the case study where each risk included also exemplary countermeasures. Further, this paper includes a description of the selected cloud risks.

Internet of Things plays a significant role in multiple sectors like agriculture, manufacturing and healthcare for collecting information to automation. The collected information is in different diversity and consists of confidential and non-confidential information. Secure handling of confidential data is a crucial task in cloud computing like storage, access control and monitoring. The blockchain based storage technique provides immutable data storage, efficient access control and dynamic monitoring to confidential data. Thus, the secure internet of things data storage, access control and monitoring using blockchain technique is proposed in this work. The patients health information that are in different formats are pruned by a decision tree algorithm and it classifies the confidential data and non-confidential data by the fuzzy rule classification technique. Depending on data owner's willing, the fuzzy rule is framed and the confidential and non-confidential data collected by internet of things sensors are classified. To provide confidentiality to confidential data, Attribute Based Encryption is applied to confidential data and stored in an off-chain mode of blockchain instead of entire data encryption and storage. The non-confidential data is stored in a plaintext form in cloud storage. When compared to support vector machine, K-nearest neighbor and Naive Bayes classification techniques, the proposed fuzzy rule based confidential data identification produces greater than 96

Multinomial logit model (MNL) is by far the most widely used discrete choice model that is widely used to explain or predict a choice from a set of two or more discrete alternatives. MNL operates within a framework of the random utility model (RUM) in which the utility of an alternative perceived by an individual consists of two components: systematic component and random component. The systematic component is usually defined as a linear function. However, practical decision processes involve complex considerations regarding various aspects of the alternatives and individual which cannot be adequately represented by simple linear models. To overcome the weakness of linear utility model and improve the performance of MNL, in this paper, we propose a general deep multinomial logit model (GDMNL) that takes advantage of both traditional MNL and deep learning. In this model, deep neural networks are applied to extend MNL by learning different nonlinear utility functions of various alternatives. The empirical study in the domain of transit route choice analysis demonstrates the validity and superiority of the proposed model.

Most research on machine learning tasks relies on the availability of true labels immediately after making a prediction. However, in many cases, the ground truth labels become available with a non-negligible delay. In general, delayed labels create two problems. First, labelled data is insufficient because the label for each data chunk will be obtained multiple times. Second, there remains a problem of concept drift due to the long period of data. In this work, we propose a novel incremental ensemble learning when delayed labels occur. First, we build a sliding time window to preserve the historical data. Then we train an adaptive classifier by labelled data in the sliding time window. It is worth noting that we improve the TrAdaBoost to expand the data of the latest moment when building an adaptive classifier. It can correctly distinguish the wrong types of source domain sample classification. Finally, we integrate the various classifiers to make predictions. We apply our algorithms to synthetic and real credit scoring datasets. The experiment results indicate our algorithms have superiority in delayed labelling setting.

Unified Modeling Language (UML) 2.0 Sequence Diagrams (UML 2.0 SD) are used to describe interactions in software systems. These diagrams must be verified in the early stages of software development process to guarantee the production of a reliable system. However, UML 2.0 SD lack formal semantics as all UML specifications, which makes their verification difficult, especially if we are modeling a critical system where the automation of verification is necessary. Communicating Sequential Processes (CSP) is a formal specification language that is suited for analysis and has many automatic verification tools. Thus, UML and CSP have complementary aspects, which are modeling and analysis. Recently, a formalization of UML 2.0 SD using CSP has been proposed in the literature; however, no automation of that formalization exists. In this paper, we propose an approach on the basis of the above formalization and a visual modeling tool to model and automatically transform UML 2.0 SD to CSP ones; thus, the existing CSP model checker can verify them. This approach aims to use UML 2.0 SD for modeling and CSP and its tools for verification. This approach is based on graph transformation, which uses AToM3 tool and proposes a metamodel of UML 2.0 SD and a graph grammar to perform the mapping of the latter into CSP. Failures-Divergence Refinement (FDR) is the model checking tool used to verify the behavioral properties of the source model transformation such as deadlock, livelock and determinism. The proposed approach and tool are illustrated through a case study.

This paper proposes a deadlock prevention policy for flexible manufacturing systems (FMSs) based on a think-globally-act-locally approach and a structurally minimal method. First, by using the think-globally-act-locally approach, a global idle place is temporarily added to a Petri net model with deadlocks. Then, at each iteration, an integer linear programming problem is formulated to design a minimal number of maximally permissive control places. Therefore, a supervisor with a low structural complexity is obtained since the number of control places is greatly compressed. Finally, by adding the designed supervisor, the resulting net model is optimally or near-optimally controlled. Three examples from the literature are used to illustrate the proposed method.

Phishing attacks are among the most common malicious activities on the Internet. During a phishing attack, cybercriminals present themselves as a trusted organization or individual. Their goal is to lure people to enter their private information, such as passwords and bank card numbers, while believing that nothing malicious is happening. The attack often starts with a phishing email, which is an email that is very similar to a legitimate email, but usually contains links to malicious websites or uses some other techniques to mislead victims. To prevent phishing attacks, it is crucial to detect phishing emails and remove them from email inbox folders. In this paper, a neural network based phishing email detection model is proposed. In comparison to some earlier approaches, our model does not use manually engineered input features. It learns character and word embeddings directly from email texts, and uses them to extract local and global features using convolutional and recurrent layers, respectively. Our model is tested on the two commonly used datasets for phishing email detection, the SpamAssassin Public Corpus and Nazario Phishing Corpus, and it achieves an accuracy of 99.81

Today's societies have become more dependent on social networks in terms of communications and interactions. These networks contain most of the people's activities, which can be public or even personal events. In the last decade, social networks have turned into more prominent platforms in managing and organizing public events. The Egyptian revolution in 2011 and the Ukrainian revolution in 2014 are good reflections of such events. However, it is not known how much the privacy issue of users is revealed in the reality as a consequence of their online interactions. In this work, we investigate the privacy issue in online social networks and its reflection on real life. Our dataset was extracted from the Facebook groups/pages that were involved in the 2019 Iraqi October revolution. Our approach generates a static network using the collected dataset. Then, we investigate the generated static network in terms of detecting potential anomalies. After that, we project the static network (including its characteristics) into a dynamic environment and generate a dynamic network for investigating the privacy issue in the real life. The contribution of this work lies in projecting a real-world static network into a dynamic environment aiming at investigating users' privacy in the real world. Finally, this kind of approach has not been given enough attention in the literature and it is therefore deeply investigated in this article.

Segmentation of medical images is a necessity for the development of healthcare systems, particularly for illness diagnosis and treatment planning. Recently, convolutional neural networks (CNNs) have gained amazing success in automatically segmenting medical images to identify organs or lesions. However, the majority of these approaches are incapable of segmenting objects of varying sizes and training on tiny, skewed datasets, both of which are typical in biomedical applications. Existing solutions use multi-scale fusion strategies to handle the difficulties posed by varying sizes, but they often employ complicated models more suited to broad semantic segmentation computer vision issues. In this research, we present an end-to-end dual-branch split architecture RGN-Net that takes the benefits of the two networks into greater account. Our technique may successfully create long-term functional relationships and collect global context data. Experiments on Lung, MoNuSeg, and DRIVE reveal that our technique reaches state-of-the-art benchmarks in order to evaluate the performance of RGN-Net.