Drones or unmanned aerial vehicles (UAVs) are increasingly deployed for disaster relief, from aiding in drug or supply delivery to search and rescue victims. For a large-scale mission, the more drones are used, the more difficult they are to control and manage in a timely manner. Automated control is one of the central techniques used in knowledge-based drone management systems to enhance the quality and efficiency of the disaster relief mission. Although many drone management systems have been developed, automatic control of these systems remains a difficult problem. Many approaches have "human in the loop" that trades the degree of autonomy with the efficiency and perhaps accuracy. Our goal is not to fully automate the system but to reduce user interactions where intelligible (e.g., selecting appropriate drones/actions based on prioritized needs) by means of significant advances in control mechanisms. This paper presents a knowledge-based framework for intelligent drone management by explicitly representing control knowledge for use in selecting appropriate task-specific actions. The paper describes how control decisions are derived using domain knowledge relevant to drones and disaster relief. By employing a blackboard control architecture, our framework provides adaptability for dynamic control behaviors and flexibility to handle unanticipated situations during the drone mission activities. The paper illustrates the use of the framework on specific scenarios.

With advances in AI (Artificial Intelligence) and the surge of Big Data, Deep learning (DL) has a remarkable impact on businesses in various domains including finance, marketing, healthcare, manufacturing, and agriculture. From sales prediction to detection and monitoring customer behaviors, its use in business analytics is proven to help grow and improve business functions. However, execution of DL can instigate high cost during both model training and deployment. Time and energy are two essential cost factors of DL computation as they imply not only monetary but also environmental consequences. Moreover, different DL architecture incurs different costs. Selecting the best model to train and deploy is hence important to the business. Most existing work focuses on building a DL model that obtains the highest accuracy or similar measures to evaluate quality of data analytics. In this paper, we argue that in practice, such performance criteria may not be adequate, for example, the most accurate model may not be the most cost-effective. This paper studies a practical approach that includes energy consumption during training and deploying the trained DL in deciding what DL model should be deployed. The paper describes how energy consumption can be estimated and illustrates the approach for cost analysis of DL application on the smart agriculture business where plant disease classification model is trained and deployed. Lessons learned from the study is included. In summary, our contributions are: (1) an approach to analyze cost of using DL model in terms of time, accuracy and energy consumption to support model selection, and (2) application of the approach on plant disease classification in smart agriculture business.

The ability to capture common characteristics among complex multi-variate time series variables can profoundly impact big data analytics in uncovering valuable insights into the relationships among them and enabling a dimensionality reduction technique. Two widely used data displays include time series and scatter plots. While the former focuses on the dynamics over time, the latter deals with static relationships among variables. Motivated by these distinctive perspectives, our research aims to maximally utilize the information captured by both at the same time. This paper presents NetScatter, a visual analytic approach to characterizing changes of pairwise relationships in a high-dimensional time series. Unlike most traditional techniques that employ a single perspective of the visual display, our approach combines static perspectives of two variables in multi-variate time series into a single representation by comparing all data instances over two different time steps. The paper also introduces a list of visual features of the representation to capture how overall data evolve. We have implemented a web-based prototype that supports a full range of operations, such as ranking, filtering, and details on demand. The paper illustrates the proposed approach on data of various sizes in different domains to demonstrate its benefits.

We aimed to understand whether warning message design recommendations address the reasons why non-experts choose to not protect themselves from cybersecurity threats. Toward that end, we synthesized literature to investigate why non-experts choose to not protect themselves, and catalog design recommendations aimed at influencing how non-experts think about threats. We then evaluated whether those recommendations addressed non-experts’ reasons. We are the first to synthesize and compare these important literatures. Our results revealed that current recommendations do not adequately address many of non-experts’ reasons for not protecting themselves. Therefore, implementing those recommendations probably will not convince those non-experts to protect themselves, which may partially explain why warning messages that implement current recommendations improve user compliance but to levels that are still lower than desired. Our results also highlight the need for future research that could lead to new warning message design recommendations that better address non-experts’ reasons for not protecting themselves.

The giant network of Internet of Things establishes connections between smart devices and people, with protocols to collect and share data. While the data is expanding at a fast pace in this era of Big Data, there are growing concerns about security and privacy policies. In the current Internet of Things ecosystems, at the intersection of the Internet of Things, Big Data, and Cybersecurity lies the subject that attracts the most attention. In aiding users in getting an adequate understanding, this paper introduces HackerNets, an interactive visualization for emerging topics in the crossing of IoT, Big Data, and Cybersecurity over time. To demonstrate the effectiveness and usefulness of HackerNets, we apply and evaluate the technique on the dataset from the social media platform.

Water is a critical natural resource that has significant impacts on human living and society. Growing population and energy consumption exacerbate the scarcity of water and our ability to manage this resource. This demonstration paper presents WaterScope, a smart and connected platform for water resource management, which integrates multiple data sources such as water level data, social media data, and water related articles. Furthermore, the tool enables forecasting underground water levels, identifying water concerns, sharing knowledge and expertise among stakeholders, and thus bringing new insights to our understanding and insights of the water supplies and resource management. The prototype engages water stakeholders who face problems of similar nature but deal with the problem in an ad-hoc and isolated manner. The interactive WaterScope platform targets creating an interconnected virtual community that aims to improve water supply resilience.

Twitter is a popular social network platform where users can interact and post texts of up to 280 characters called tweets. Hashtags, hyperlinked words in tweets, have increasingly become crucial for tweet retrieval and search. Using hashtags for tweet topic classification is a challenging problem because of context dependent among words, slangs, abbreviation and emoticons in a short tweet along with evolving use of hashtags. Since Twitter generates millions of tweets daily, tweet analytics is a fundamental problem of Big data stream that often requires a real-time Distributed processing. This paper proposes a distributed online approach to tweet topic classification with hashtags. Being implemented on Apache Storm, a distributed real time framework, our approach incrementally identifies and updates a set of strong predictors in the Naïve Bayes model for classifying each incoming tweet instance. Preliminary experiments show promising results with up to 97% accuracy and 37% increase in throughput on eight processors.

Smart Grid is an important cyber-physical infrastructure for power generation, transmission and distribution. As energy becomes a necessity for our modern living, effective management of Smart Grid can have great impact on power services. Recent smart meter technology has provided an opportunity that allows an economic model of utility to be more flexible and optimal. Construction of such a model requires better understanding of consumer behaviors in response to a given pricing. Most existing research uses mathematical models that tend to rely on assumptions that may not be realistically realizable. Majority of empirical work aims to predict or optimize utility loads. To the best of our knowledge, no empirical published work has addressed this issue. Our research aims to find an appropriate Big data analytic approach to learning consumption pattern of each household at appliance level. Particularly, this paper investigates how machine learning can be used to predict the change of customer's consumption habits for lower pricing. The problem is non-trivial due to conflicting dependent factors that are dynamically changing. Our preliminary results, obtained from three popular machine-learning models using synthesized data, show relatively high accuracy ranging from 61.5% to 99.4% for various appliances. Although the results are encouraging, we provide insights towards the validity and improvement of these predictive models.

Cyber-physical systems (CPS) tightly integrate physical and computing processes by monitoring and control data interacting between them via underlying networks. Software Defined Network (SDN) Technology has increasingly become essential in many advanced computer networks, including those in modern CPS, to provide flexible and agile network development. Despite many benefits that SDN offers, malicious attacks that can eventually prevent network services are unavoidable. Among the most predominant attacks on SDN controller layer, Link Discovery Attack and ARP (Address Resolution Protocol) Spoofing Attack are fundamental in that they are the gateways of many other SDN threats and attacks. To defend these attacks, most existing techniques either rely on relatively complex data validation techniques or use thresholds that can be subjective and unable to detect more than one type of attacks at a time if one deciding factor is used. While Big data technology, particularly machine learning, has been widely used for intrusion/anomaly detection, little has been done in SDN. This paper explores how well this technology can be used to predict these SDN attacks. By employing typical machine learning algorithms on simulated data of routing in SDN when attacks occur, preliminary results, obtained from four machine learning models, show the average area under ROC curve of over 96% and 92% for sample size 50,970 (12 switches) and 60,000 (20 switches), respectively. Further experiments show near-linear scaling in training time for the best performing algorithm when sample size grows up to 100,000.

Twitter is one of the most popular microblogging platforms where users can interact with each other by posting texts of up to 140 characters called tweets. Because of the large and fast growing number of tweets being generated daily, tweet analytics is viewed as one of the fundamental problems of Big data stream. Recently, hashtags, hyperlinked words, in tweets have been applied for tweet retrieval, trend/event detection and advertisement. However, using hashtags for tweet classification remains challenging because we have to cope with context dependent words, slangs, abbreviations, and emoticons with a limited small number of words and an evolving use of hashtags. Most existing approaches deal with classifying tweet sentiments by using the lexicon and meaning of hashtags. Our research aims to classify tweets by topics. Unlike sentiment analytics, hashtags for describing a topic need to be more diverse to cover various aspects of a topic. This paper presents a tweet analytics approach that uses domain-specific knowledge to create a set of strong hashtag predictors for tweet topic classification. The paper describes the approach and preliminary experiments that show promising results toward Big data tweet analytics.

Protecting network-accessible resources from unauthorized use requires understanding of network vulnerability. Building attack models as chains of vulnerability exploits by an attacker can help locate flaws to prevent attacks. This paper presents an approach to automatically analyzing security attack models called exploit-based analysis. By using knowledge about exploitability of network vulnerabilities combined with a Markov computational process commonly used in link analysis, exploit based analysis ranks nodes in the attack model in order of their likelihoods of being compromised. The paper discusses the details of the approach, illustrates its use, and compares results with an existing technique.

The incidence of cyberbullying has been on the increase as more young people are spending time socializing online. Cyberbullying manifests itself through the sending of embarrassing pictures, mean-spirited text messages, emails and/or postings on social media. Technology has exacerbated the problem by allowing bullying to be accomplished easily and anonymously. The problem of cyberbulling has been brought at the forefront of many colleges and high schools because it is spreading. Indeed, studies show that 39% of social network users have been victims of cyberbulling, compared with 22% of online teens who do not use social networks. Additionally, 88% of teens using social media say they have witness an instance of cyberbulling. Victims of cyberbullying experience negative emotional, physical and social consequences, sometimes leading to the victim's suicide.

Social networks can provide useful information to enhance mobile cyber-physical applications. With advanced sensor and powerful mobile Internet device technologies, future social networking is likely to be cyber-physical, combining social behaviors and human interactions with computational and physical processes. This paper addresses the problem of analyzing and extracting useful properties from a social cyber-physical network. In particular, we present a graph-based approach to automatically identifying essential nodes that hold the network together called key breakers. Knowing key breakers can help us contain spread of diseases, computer viruses, and undesirable communication or security breaches in the network. The paper describes the approach and provides illustrations, empirical evaluation and a comparative study with existing techniques.

An attack graph increasingly plays an important role in network security. It shows possible paths of actions consisting of the network vulnerability exploits that can lead to security breaches. Because most attack graphs are very large and complex, much research has focused on how these graphs can be automatically and efficiently generated. However, little has been done on attack graph analysis, namely how we can use attack graphs to better protect the network. This paper addresses the latter issue. We present a suit of systematic approaches to statically analyzing attack graphs by means of reasoning mechanisms based on logical expressions and conditional preference networks. The proposed approaches are general and theoretically grounded. The paper describes the approaches in details. We show how the resulting analysis can help derive many useful decisions. For example, it can assist a security administrator in selecting most cost-effective countermeasures, based on his preference criteria, to improve the security flaws found in the attack graph. For understandability, we illustrate our approach by presenting a study of a simple and small but realistic case scenario.

This paper addresses a fundamental issue of Web service composition. We present a simple but powerful conceptual model that leads to a scalable approach to automatically constructing a composite Web service to meet its requirements by using as few services as possible. Our approach is based on a state space model that has a monotone property to allow efficient search along with efficient algorithms for pruning and simple parallelization. We provide both empirical and theoretical analyses of the proposed approach and show that it has time complexity of O(n2), for a repository with n services. However, the approach takes linear time for sequential compositions when service applicability is performed by service discovery and thus, it is shown to give asymptotically optimal performance. Although optimality in the number of services deployed is not guaranteed, our experiments on public benchmark data sets show correct optimized solutions 100% of the time, with a reduction in the average running time, compared to a well-performed planning-based system, of better than 35% over 207 composition problems.

The order in which software components are tested can have a significant impact on the number of stubs required during component integration testing. This paper presents an efficient approach that applies heuristics based on a given software component test dependency graph to automatically generate a test order that requires a (near) minimal number of test stubs. Thus, the approach reduces testing effort and cost. The paper describes the proposed approach, analyses its complexity and illustrates its use. Comparison with three well known graph-based approaches, for a real-world software application, shows that only the classic Le Traon et al.'s approach and ours give an optimal number of stubs. However, experiments on randomly simulated dependency models with 100 to 10,000 components show that our approach has a significant performance advantage with a reduction in the average running time of 96.01%.

Advances in Internet and software technology play increasing roles in creating work environments and online services that impact our everyday living. Web service computing paradigm has revolutionized how these software applications can be developed rapidly and reliably by employing available services available in standard forms across the web. Automatic composition of Web services is necessary and challenging when there are a huge growing number of available Web services but no single service offers the desired function. This paper presents an approach to automatically combining published Web services to meet the application requirements using as few services as possible. Our approach is based on heuristic algorithms that incorporate semantics of services from their ontology into identifying valid executable services in the composition structure. The paper describes the approach and evaluates its performance on public service repositories. Our experimental results show correct optimized solutions 100% of the time, with a reduction in the average running time, compared to champion of Web service chal-lenge'06 competition, of 60.2% over 31 composition problems.

Role-based access control (RBAC) is the most predominant access control model in today's security management due to its ability to simplify authorization, and flexibility to specify and enforce protection policies. In RBAC, Separation of Duty (SoD) constrains user role authorization to protect sensitive information from frauds due to conflicts of interests. SoD constraints are commonly defined by mutually exclusive roles (MER) (e.g., bank teller and auditor). This paper proposes practical computational techniques for analyzing SoD by integrating workflows of the enterprise processes into the RBAC framework. Specifically, we present 1) an algorithm for generating MER to enforce SoD, and 2) a verification algorithm to check if a given RBAC state (role authorization and user-role assignments) satisfies a given type of SoD constraint or not. The paper discusses the details of the approach and illustrates its use in a loan application domain.

Research has successfully applied model checking, a formal verification technique, to automatically generate chains of vulnerability exploits that an attacker can use to reach his goal. Due to the combinatorial explosion of the chain generation problem space, model checkers do not scale well to networks containing a large number of hosts. This paper proposes a methodology that uses a host-centric modeling approach together with a monotonicity assumption to alleviate the scalability problem of model checkers. We describe the proposed approach, its limitations, and show how it can reduce the time complexity of chain generation to a quadratic polynomial of the number of hosts, both theoretically and empirically. We also compare its advantages over similar customized graph-based approaches.

Any Web service procurement involves complex negotiations on non-functional requirements in order to obtain an integrative solution for both the provider and consumer perspectives. Software that facilitates automated Web service procurement negotiations is valuable not only for the consumers and providers to continuously bargain for their customizations and tailor their offerings, but also to discover overlooked alternative solutions and to maintain documented rationales for future references and reuse. To date, software frameworks and mechanisms that provide flexible environments to support automated negotiations in web service procurement are lagging. This paper presents a tradeoff-based automated negotiation approach to support Web service procurement. The approach employs an iterative tradeoff mechanism for evaluating opponent's offers and generating counter-offers of mutual gain based on selected quality of service parameters. The paper presents the negotiation approach in details along with an illustrated case study of negotiation in Web service procurement for a medical insurance company.

Managing access authorities is critical to the security of information systems. To prevent fraud or abuse due to conflict of interests, a well-known authorization constraint called separation of duty (SoD) is commonly applied. SoD ensures that no single user receives too many authorities. Enforcement of authorization constraints such as SoD in large organizations can be difficult due to the large number of information system users, the variety of assets involved, and tasks that require roles that may be shared or delegated at multiple levels. Most existing work in this area focuses on specifications of SoD constraints and assumes that constraints can be enforced by logical inference mechanisms at run-time. A drawback of this approach is that when violations occur, finding alternative role activations at run-time may not be feasible. This can result in delays or even failure for critical service transactions. Moreover, logic-based systems are difficult to understand and do not scale easily. This paper presents an algorithmic set-based approach that automatically checks for SoD compliance prior to run-time by searching for a set of valid role activations. The paper discusses details of this approach and illustrates its use in managing access authorizations in a health insurance claim processing system.

Assessing risk associated with safety or mission-critical embedded systems often lacks detailed considerations of software, if at all. Performing risk assessment, which considers software in its entirety, is challenging, yet necessary. This necessity stems from the need to understand the contribution of software to the risks associated with a system. We propose a modeling approach that extends dynamic flowgraph methodology via using second order decision tables to represent functions of the system, including that of the software, and the environment

Current techniques that specifically address risks associated with software are lacking. Although many risk assessment techniques have been successfully applied in various safety-related systems, they often omit detailed software properties or exclude application perspectives of an entire system. Furthermore, most existing tools require manual assessment of hazard analysis. This paper presents a model-based reasoning approach for assessment of software risks by employing knowledge models of software faults and the system in which the software is used along with reasoning mechanisms that allow automated qualitative inferences for hazard analysis. We give an example of a software controller for a reactor main feed water system to illustrate the approach.

Mutations are structural changes in DNA that can cause protein malfunction and genetic disease. This paper describes a machine learning approach to analyzing mutations associated to Osteogenesis Imperfecta (OI), also known as brittle bone disease. We apply SORCER, a second-order rule induction system to predict clinical phenotypes of OI from mutation and neighboring amino acid sequences in the COLIA1 gene. On the average, over ten 10-fold cross-validations, SORCER gives more accurate results than C4.5 with average accuracy of about 81.2%. The paper discusses the advantages and limitations of SORCER and demonstrates its use to provide initial exploration of biological sequences.

The success of data mining techniques can be measured by the usefulness of models they produce. Often these models must be explainable as well as accurate. While decision tables are easy to interpret and explain to virtually all users, there has been little study of whether such simple models are powerful enough to use for data mining. This paper presents SORCER, a learning system that induces second-order decision tables from a given data set. Second-order tables are database relations in which rows have sets of atomic values as components. Using sets of values interpreted as disjunctions, secondorder tables provide simple and compact representations that both enhance comprehensibility and facilitate efficient management. Unlike many other learning systems, to further promote comprehensibility, SORCER attempts to generate a minimum number of rows. SORCER’s induction algorithm can be viewed as a table compression technique in which a traditional table, representing training data, is transformed into a second-order table with fewer rows by merging rows in ways that preserve consistency with the training data. We compare SORCER to three classification systems: C4.5, CBA and a Naive Bayesian classifier. Experimental results on 26 data sets show that the average error rate obtained from SORCER, using a simple compression method, is lower than those of the Naive Bayesian classifier and C4.5 and is competitive with CBA’s average error rate. Using a slightly more sophisticated compression, SORCER’s average error rate is the lowest of all.

Investigates a machine learning approach to discovering predictive relationships that can be used to integrate solar and ocean-atmospheric conditions into forecasts of regional water flows. In particular, we apply decision-tree learning and a recently developed inductive technique called "second-order table compression" to generate predictive models of future water inflows of Lake Okeechobee, a primary source of water supply for south Florida, USA. We describe SORCER (Second-Order Relation Compression for Extraction of Rules), which is a 2nd-order table compression learning system, and compare its results with those obtained from a well-established decision tree learner, C4.5. On the average, in ten 10-fold cross-validations, SORCER obtained a slightly lower error rate than C4.5. We discuss the implications of these results.

This paper addresses the problem of integrating the effects of climate history and solar variability, to enhance regional hydrologic forecasting using neural networks. A previous attempt at modeling the inflow to Lake Okeechobee employed a multilayered perceptron (see Trimble et al, 1998). While the resulting model was able to capture some regularities of the measured inflow, it was far from being a useful predictive model. We continue the lake inflow modeling effort by examining data representation, quadratic input transformations, and time-delay neural networks.

Achieving efficient software implementations requires a great deal of knowledge, intelligence, and expertise on the part of programmers. One way to enhance software productivity is to incorporate the knowledge and skills of expert programmers into software synthesis systems to automate software development processes. Although many software synthesis systems have been developed, automatic control of synthesis remains a difficult problem. Understanding the role of expertise in software synthesis, and making it more explicit, can help us not only to gain autonomy in controlling the synthesis processes but also to better justify the design, implementations, selection of data structures or algorithms employed in constructing code. Our project aims at making synthesis as autonomous as possible by advances in intelligent control mechanisms to reduce user interaction in the synthesizer. In our earlier work, a blackboard control framework for controlling synthesis processes was introduced. This paper describes how the control framework language was designed and how knowledge in the knowledge bases of the framework was acquired and constructed. We present an example that shows how programming expertise can be used to increase the degree of autonomy in synthesis control, in particular by automating the selection of an appropriate data structure implementation.

The blackboard model is a versatile framework for implementing expert systems. In recent years the speed of blackboard access has been improved but the execution cycle, which manages knowledge source activation and execution, still needs substantial improvement. A method of condensing knowledge sources is presented, and it is shown how to efficiently activate and execute a class of blackboard application systems using the condensed representation. It is shown that the time complexity of the execution cycle of a condensed system is faster than that of the conventional approach by the ratio of the time required for blackboard retrievals to the time required for vector element retrievals. In practice, this ratio is approximately four order of magnitude.>

Effective monitoring of device-supported patients in the intensive care unit (ICU) is complex, involving interpretation of many variables, comparative evaluation of many therapy options, and control of many patient-management parameters. Even skilled clinicians make errors that limit the quality of care, harm patients, or cause life-threatening situations. A growing body of research aims to improve ICU monitoring with computer technology. Most of this research falls in two areas: (a) short-term engineering of practical solutions to narrowly defined immediate problems (e.g. smart alarm systems); or (b) basic research on fundamental issues potentially relevant to ICU monitoring (e.g. temporal reasoning). By contrast, our project aims to develop a more comprehensive ‘intelligent agent’, having a broad range of capabilities, to cooperate on the ICU team. We do not aim to produce a practical system suitable for near-term deployment in the ICU, but rather a ‘proof of concept’, an experimental system that: (a) demonstrably performs and coordinates a range of intelligent reasoning tasks of use in ICU monitoring; (b) does so reliably in a significant range of medical situations; and (c) arguably will scale up to meet the comprehensive set of practical requirements with an appropriate development effort. We have developed an experimental system called Guardian, which exhibits several of the required capabilities and utilizes an underlying architecture hypothesized to support the full range of required capabilities. In this paper, we describe the Guardian system, its architecture, and its current knowledge base. We describe its performance and summarize the results of preliminary evaluations. Finally, we discuss ongoing and planned research on Guardian.