Intelligent Distributed Computing X

An approach for adapting distributed applications in response to changes in user requirements and resource availability is presented. The notion of elasticity enables capabilities and resources to be dynamically provisioned and released. However, existing applications do not inherently support elastic capabilities and resources. To solve this problem, we propose two novel functions: dynamic deployment of components and dividing and merging components. The former enables components to relocate themselves at new servers when provisioning the servers and at remaining servers when deprovisioning servers, while the latter enables the states of components to be divided, passed to other components, and merged with other components in accordance with user-defined functions. We constructed a middleware system for Java-based general-purpose software components with the two functions because they are useful to adapt applications to elasticity in cloud computing. The proposed system is useful because it enables applications be operated with elastic capabilities and resources in cloud computing.

In Service Computing (SC), online Semantic Web services (SWs) is evolving over time and the increasing number of SWs with the same function on the Internet, a great amount of candidate services emerge. So, efficiency and effectiveness has become a stern challenge for distributed discovery to tackle uniformed behavior evolution of service and maintain high efficiency for large-scale computing. The distributed discovery of SWs according to their functionality increases the capability of an application to fulfill their own goals. In this paper, we describe an efficient and an effective approach for improving the performance and effectiveness of distributed discovery of SWs in P2P systems. As most Web services lack a rich semantic description, we extend the distributed discovery process by exploiting collaborative ranking to estimate the similarity of a SWs being used by existing hybrid matching technique of OWL-S (Ontology Web Language for Services) process models in order to reduce costs and execution time. We mapped our distributed discovery of OWL-S process models by developing a real application based on Gamma Distribution; a technique used to decrease the bandwidth consumption and to enhance the scalability of P2P systems. The particularity of the Gamma Distribution is then integrated for disseminating request about the P2P networks to perform quality based ranking so that the best SWs can be recommended first. The experimental result indicates that our approach is efficient and able to reduce considerably the execution time and the number of message overhead, while preserving high levels of the distributed discovery of SWs on large-size P2P networks.

In this paper we propose a framework based on BDI software agents for the modeling and simulation of dynamic systems. The target system is broken down into a number of interacting components. Each component is then mapped to a BDI agent that captures its behavioral aspects. The system model is described as a multi-agent program that is specified using the Jason agent-oriented programming language.

This paper presents an approach for the configuration, deployment and monitoring of distributed applications in a smart environment. This approach takes into consideration the heterogeneity and the dynamicity of such environments and deals with resource privacy. We propose to describe the available hardware infrastructure and the deployable applications using graphs, and provide a mathematical formalisation of the deployment process based on graph homomorphisms. A decentralised version of a branch and bound graph-matching algorithm is used to find the available hardware entities of the infrastructure that can be used to run the application, respecting its requirements. At last, we describe a goal-directed Multi-Agent System (MAS) for the deployment of applications in ambient systems. We show that the multi-agent paradigm is well-adapted to provide a clear separation between the applicative and the hardware layers, thus increasing resource privacy.

In this paper, we address ambient systems whose computational process is based on autonomous and context-aware intelligent agents. The planning management framework we propose is agent-centered and looks for an efficient guidance improving the satisfaction of the agent’s intentions, taken as a whole. This also includes the intentions pushed by some opportunistic situations, according to their relevances for the agents. Originally, the formal model we propose allows to dynamically schedule the plans of the intentions concurrently, allowing to extract the traces having a maximum relevance coding.

The paper presents a visual language that can help users of a context-aware application represent the current situation, or situations they wish detected, in a language that is both formally defined, and readable and understandable by humans and machines alike. Inspired from Regular Expressions, the concept of Extended Concept Pattern provides both conciseness and expressive power, allowing for specifying negation, and for indicating repeating or alternative structures.

The Internet of Things (IoT) unleashes great opportunities to improve our way of living and working through a seamless and highly dynamic cooperation among heterogeneous things including both computer-based systems and physical objects. However, properly dealing with the design, development, deployment and runtime management of IoT applications means to provide solutions for a multitude of challenges related to intelligent distributed systems within the IoT. In this paper we propose Model-Driven Engineering (MDE) as a keyenabler for applications running on intelligent distributed IoT systems. MDE helps in tackling challenges and supporting the lifecycle of such systems. Specifically, we introduce MDE4IoT, an MDE approach enabling the modelling of things and supporting intelligence as self-adaptation of Emergent Configurations in the IoT. Moreover, we show how MDE, and in particular MDE4IoT, can help in tackling several challenges by providing the Smart Street Lights concrete case.

Methods and algorithms for detection of traffic anomalies in multi-service networks play a key role in creating the malware intrusion detection and prevention systems in modern communication infrastructures. The major requirement imposed to such systems is the ability to find anomalies and, respectively, intrusions in real time. Complexity of this problem is caused in many ways by incompleteness, discrepancy and variety of distribution laws at streams in a multi-service traffic. The paper represents a new technique for traffic anomaly detection in multiservice networks. It is based on using modified adaptation algorithms without identification and fuzzy logical inference rules. Results of an experimental assessment of the technique are discussed.

Nowadays, Role-Based Access Control (RBAC) is a widespread access control model. Search of the “users-roles” and “roles-permissions” mappings for the given “users-permissions” mapping is the problem of Data Mining called as Role Mining Problem (RMP). However, in the known works devoted to the RMP, the problem of RBAC scheme reconfiguration and methods for solving it are not considered. The paper defines the statement of the problem of RBAC scheme reconfiguration and suggests a genetic algorithm for solving it. Experimental results show the algorithm has a high enough effectiveness.

Android platforms are known as the less security smartphone devices. The increasing number of malicious apps published on Android markets suppose an important threat to users sensitive data, compromising more devices everyday. The commercial solutions that aims to fight against this malware are based on signature methodologies whose detection ratio is low. Furthermore, these engines can be easily defeated by obfuscation techniques, which are extremely common in app plagiarism. This work aims to improve malware detection using only the binary information and the permissions that are normally used by the anti-virus engines, in order to provide a scalable solution based on machine learning. In order to evaluate the performance of this approach, we carry out our experiments using 5000 malware and 5000 benign-ware, and compare the results with 56 Anti-Virus Engines from VirusTotal.

It is largely recognized that many missions may be easily performed by unmanned vehicles both in military and in civil domain. Literature shows a large inventory of their applications with operational and logistical challenges. Comparing different types of missions, a multi-vehicle approach is able to guarantee better performances and minimum costs, as long as they are coordinated. Thus, the problem to guarantee the better platform configuration to perform the mission becomes architecting the best fleet. This paper proposes an approach to identify the best fleet to perform an envelope of missions, by transforming the architecting activity in an optimization problem. A classification of unmanned vehicles missions and the formal definition of the problem are proposed.

We introduce Spatial Tuples, an extension of the basic tuplebased model for distributed multi-agent system coordination where (i) tuples are conceptually placed in the physical world and possibly move, (ii) the behaviour of coordination primitives may depend on the spatial properties of the coordinating agents, and (iii) the tuple space can be conceived as a virtual layer augmenting physical reality. Motivated by the needs of mobile augmented-reality applications, Spatial Tuples explicitly aims at supporting space-aware and space-based coordination in agent-based pervasive computing scenarios.

Swarm robotics is focused on creating intelligent systems from large number of simple robots. The majority of nowadays robots are bound to operations within mono-modal locomotion (i.e. land, air or water). However, some animals have the capacity to alter their locomotion modalities to suit various terrains, operating at high levels of competence in a range of substrates. One of the most significant challenges in bio-inspired robotics is to determine how to use multi-modal locomotion to help robots perform a variety of tasks. In this paper, we investigate the use of multi-modal locomotion on a swarm of robots through a multi-target search algorithm inspired from the behavior of ying ants. Features of swarm intelligence such as distributivity, robustness and scalability are ensured by the proposed algorithm. Although the simplicity of movement policies of each agent, complex and efficient exploration is achieved at the team level.

Observing the features of the information actually stored in the Web, we can recognize that an important issue to be investigated is that of discovering relationships between groups of objects. In particular, a great interest is emerging on finding groups of objects mutually linked by reciprocal relationships of trustworthiness. In this paper, we propose a model to represent the case of trust-based groups of objects, and we present an algorithm for detecting trust associations in virtual communities in presence of these groups. Such an algorithm consists in determining particular sub-structures of the community, called trust groups, representing objects mutually connected by strong trust relationships. We technically formalize our idea and algorithm, and we present a complete example of how our approach works.

Attention is a very important cognitive and behavioral process, by means of which an individual is able to focus on a single aspect of information, while ignoring others. In a time in which we are drawn in notifications, beeps, vibrations and blinking messages, the ability to focus becomes increasingly important. This is true in many different domains, from the workplace to the classroom. In this paper we present a non-intrusive distributed system for monitoring attention in teams of people. It is especially suited for teams working at the computer. The presented system is able to provide real-time information about each individual as well as information about the team. It can be very useful for team managers to identify potentially distracting events or individuals, as well as to detect the onset of mental fatigue or boredom, which significantly influence attention. In the overall, this tool may prove very useful for team managers to implement better human resources management strategies.

This paper analyses a way to cope with incomplete information, namely information regarding the conflict style used by parties. This analysis is important because it enables us to develop a more accurate and informed conflict style classification method to promote better strategies. To develop this proposal, an experiment using a combination of Bayesian Networks with Speculative Computation is depicted. Thus, in this work, was firstly identified and applied a set of methods for classifying conflict styles with incomplete information; secondly, the approach was validating opposing data collected from a web-based negotiation game. From the experiment outcomes, we can concluded that it is possible to cope with incomplete information by producing valid conflict style default values and, particularly, to anticipate competing postures through the dynamic generation of recommendations for a conflict manager. The findings suggest that this approach is suitable for handling incomplete information in this context and can be applied in a viable and feasible way.

Online Social Networks are suitable environments for e-Learning for several reasons. First of all, there are many similarities between social network groups and classrooms. Furthermore, trust relationships taking place within groups can be exploited to give to the users the needed motivations to be engaged in classroom activities. In this paper we exploit information about users’ skills, interactions and trust relationships, which are supposed to be available on Online Social Networks, to design a model for managing formation and evolution of e-Learning classes and providing suggestions to a user about the best class to join with and to the class itself about the best students to accept. The proposed approach is validated by a simulation which proves the convergence of the distributed algorithm discussed in this paper.

The problem of intelligent Grid computing and job-flow scheduling with regard to preferences given by various groups of virtual organization (VO) stakeholders (such as users, resource owners and administrators) is studied. A specific flexible resources share algorithm is proposed for job-flow scheduling which enables to achive a balance between the VO stakeholders’ conflicting preferences and policies. This approach provides greater VO scheduling fairness, improves the overall quality of service and resource load efficiency. Two different metrics are introduced to find a scheduling solution balanced between VO stakeholders. Experimental results prove that the cyclic scheduling scheme allows establishing efficient cooperation between different VO stakeholders even if their goals and preferences are contradictory.

Since the advent of Telecommunication networks in the early 60’s, routing has become a recurrent problem with evergrowing complexity due to the simultaneous share of resources, stringent Quality of Service (QoS) constraints and unmanageable network scales (size, speed and exchanged data volume) by conventional route finding schemes. This paper considers a particular class of routing problems where the route to be found needs to simultaneously fulfill different requirements in terms of e.g. maximum latency, loss rate or any other cost measure. The manuscript delves into the application of the Coral Reefs Optimization and the Firey Algorithm, two of the latest bio-inspired meta-heuristic techniques reported to outperform other approximative solvers in a wide range of optimization scenarios. Results obtained from Monte Carlo simulations over synthetic network instances will shed light on the comparative performance of these two algorithms, with emphasis on their convergence speed and statistical significance.

Ordering nodes by rank is a benchmark used in several contexts, from recommendation-based trust networks to e-commerce, search engines and websites ranking. In these scenarios, the node rank depends on the set of links the node establishes, hence it becomes important to choose appropriately the nodes to connect to. The problem of finding which nodes to connect to in order to achieve the best possible rank is known as the best attachment problem. Since in the general case the best attachment problem is NP-hard, in this work we propose heuristics that produce near-optimal results while being computable in polynomial time; simulations on different networks show that our proposals preserve both effectiveness and feasibility in obtaining the best rank.

The recent advent of ‘Internet of Things’ technologies is set to bring about a plethora of heterogeneous data sources to our immediate environment. In this work, we put forward a novel concept of dynamic intelligent virtual sensors (DIVS) in order to support the creation of services designed to tackle complex problems based on reasoning about various types of data. While in most of works presented in the literature virtual sensors are concerned with homogeneous data and/or static aggregation of data sources, we define DIVS to integrate heterogeneous and distributed sensors in a dynamic manner. This paper illustrates how to design and build such systems based on a smart building case study. Moreover, we propose a versatile framework that supports collaboration between DIVS, via a semantics-empowered search heuristic, aimed towards improving their performance.

Crowd sourcing project enable the use of community of users towards the benefit of society. Aligned with trends such as smart city design and the internet of things the range of application are only restricted by human imagination. Taking the case of urban driving, it is already possible to estimate roadblocks, congestions and issue real-time alerts to users using popular applications. The approach taken in this papers, furthers this analysis by providing means to analyse the route cause of not only such events but also dangerous driving habits from users. Making use of machine learning algorithms, big data and distributed systems, a work-flow based on the PHESS Driving platform was developed. Results achieved are satisfactory in the field tests produced, giving reason to some popular common sense, as well as, new theories for dangerous driving events.

In the last decade the interest in adaptive models for non-stationary environments has gained momentum within the research community due to an increasing number of application scenarios generating non-stationary data streams. In this context the literature has been specially rich in terms of ensemble techniques, which in their majority have focused on taking advantage of past information in the form of already trained predictive models and other alternatives alike. This manuscript elaborates on a rather different approach, which hinges on extracting the essential predictive information of past trained models and determining therefrom the best candidates (intelligent sample matchmaking) for training the predictive model of the current data batch. This novel perspective is of inherent utility for data streams characterized by short-length unbalanced data batches, situation where the so-called trade-off between plasticity and stability must be carefully met. The approach is evaluated on a synthetic data set that simulates a non-stationary environment with recurrently changing concept drift. The proposed approach is shown to perform competitively when adapting to a sudden and recurrent change with respect to the state of the art, but without storing all the past trained models and by lessening its computational complexity in terms of model evaluations. These promising results motivate future research aimed at validating the proposed strategy on other scenarios under concept drift, such as those characterized by semi-supervised data streams.