Thesis (M.S. in Information Systems)--Naval Postgraduate School, March 1992. / Thesis Advisor: Bhargava, Hemant. "March 1992." Description based on title screen as viewed on March 4, 2009. Includes bibliographical references (p. 71-72). Also available in print.

Thesis (Ph.D.)--University of Delaware, 2006. / Principal faculty advisor: Busby N. O. Attoh-Okine, Dept. of Civil & Environmental Engineering. Includes bibliographical references.

Thesis (Ph. D.)--West Virginia University, 2004. / Title from document title page. Document formatted into pages; contains x, 118 p. : ill. (some col.). Includes abstract. Includes bibliographical references (p. 104-118).

Intrusion detection is the essential part of network security in combating against illegal network access or malicious cyberattacks. Due to the constantly evolving nature of cyber attacks, it has been a technical challenge for an intrusion detection system (IDS) to effectively recognize unknown attacks or known attacks with inadequate training data. Therefore in this dissertation work, an innovative two-stage classifier is developed for accurately and efficiently detecting both unknown attacks and known attacks with insufficient or inaccurate training information. The novel two-stage fuzzy classification scheme is based on advanced machine learning techniques specifically for handling the ambiguity of traffic connections and network data. In the first stage of the classification, a fuzzy C-means (FCM) algorithm is employed to softly compute and optimize clustering centers of the training datasets with some degree of fuzziness counting for feature inaccuracy and ambiguity in the training data. Subsequently, a distance-weighted k-NN (k-nearest neighbors) classifier, combined with the Dempster-Shafer Theory (DST), is introduced to assess the belief functions and pignistic probabilities of the incoming data associated with each of known classes to further address the data uncertainty issue in the cyberattack data. In the second stage of the proposed classification algorithm, a subsequent classification scheme is implemented based on the obtained pignistic probabilities and their entropy functions to determine if the input data are normal, one of the known attacks or an unknown attack. Secondly, to strengthen the robustness to attacks, we form the three-layer hierarchy ensemble classifier based on the FCM weighted k-NN DST classifier to have more precise inferences than those made by a single classifier. The proposed intrusion detection algorithm is evaluated through the application of the KDD’99 datasets and their variants containing known and unknown attacks. The experimental results show that the new two-stage fuzzy KNN-DST classifier outperforms other well-known classifiers in intrusion detection and is especially effective in detecting unknown attacks.

WICKRAMARATHNE, T. L. (M.S., Electrical and Computer Engineering) A Belief Theoretic Approach for Automated Collaborative Filtering (May 2008) Abstract of a thesis at the University of Miami. Thesis supervised by Professor Kamal Premaratne. No. of pages in text. (84) Automated Collaborative Filtering (ACF) is one of the most successful strategies available for recommender systems. Application of ACF in more sensitive and critical applications however has been hampered by the absence of better mechanisms to accommodate imperfections (ambiguities and uncertainties in ratings, missing ratings, etc.) that are inherent in user preference ratings and propagate such imperfections throughout the decision making process. Thus one is compelled to make various "assumptions" regarding the user preferences giving rise to predictions that lack sufficient integrity. With its Dempster-Shafer belief theoretic basis, CoFiDS, the automated Collaborative Filtering algorithm proposed in this thesis, can (a) represent a wide variety of data imperfections; (b) propagate the partial knowledge that such data imperfections generate throughout the decision-making process; and (c) conveniently incorporate contextual information from multiple sources. The "soft" predictions that CoFiDS generates provide substantial exibility to the domain expert. Depending on the associated DS theoretic belief-plausibility measures, the domain expert can either render a "hard" decision or narrow down the possible set of predictions to as smaller set as necessary. With its capability to accommodate data imperfections, CoFiDS widens the applicability of ACF, from the more popular domains, such as movie and book recommendations, to more sensitive and critical problem domains, such as medical expert support systems, homeland security and surveillance, etc. We use a benchmark movie dataset and a synthetic dataset to validate CoFiDS and compare it to several existing ACF systems.

Thesis (M.S.)--California Polytechnic State University, 2009. / Title from PDF title page; viewed on June 11, 2009. "June 2009." "In partial fulfillment of the requirements for the degree [of] Master of Science in Electrical Engineering." "Presented to the Electrical Engineering faculty of California Polytechnic State University, San Luis Obispo." Major professor: John Saghri, Ph.D. Includes bibliographical references (p. 72-74). Also available on microfiche.

Thesis (M.S.)--West Virginia University, 2005. / Title from document title page. Document formatted into pages; contains ix, 50 p. : ill. (some col.). Includes abstract. Includes bibliographical references (p. 48-50).

Policy-makers of today's public transportation investment projects engage in debates in which the reasonableness and clarity of their judgment are tested many times. How to recommend the transportation system that achieves project's goals and different stakeholders' needs in a most logical and justifiable manner is the main question of this dissertation. This study develops a new decision-making approach, Belief Reasoning method, for evaluating public transportation systems in the planning process. The proposed approach applies a reasoning map to model how experts perceive and reason transportation alternatives to lead to the project's goals. It applies the belief measures in the Dempster-Shafer theory of evidence as the mathematical mechanism to represent knowledge under uncertainty and ambiguity and to analyze the degree of achievement of stated goals. Three phases are involved in implementing the Belief Reasoning method. First, a set of goals, a set of characteristics of the alternatives, a set of performances and impacts are identified and the reasoning map, which connects the alternatives to the goals through a series of causal relations, is constructed. Second, a knowledge base is developed through interviewing the experts their degree of belief associated with individual premises and relations, and then aggregating the expert opinions. Third, the model is executed and the results are evaluated in three ways: (i) the transportation alternatives are evaluated based on the degree of belief for achieving individual goals; (ii) the integrity of the reasoning process is evaluated based on the measures of uncertainty associated with information used; and (iii) the critical reasoning chains that significantly influence the outcome are determined based on the sensitivity analysis. The Belief Reasoning method is compared with the Bayesian reasoning, which uses the probability measures as the measure of uncertainty. Also it is compared with the Analytical Hierarchy Process method, which uses a hierarchical tree structure and a weighting scheme. The numerical examples in transit planning are developed for comparison. The proposed Belief Reasoning method has advantages over these traditional evaluation and reasoning methods in several ways. • Use of a reasoning map structure together with an inference process, instead of a tree structure together with a weighting scheme, allows modeling interdependency, redundancy and interactions among variables, usually found in transportation systems. • Use of belief measures in Dempster-Shafer theory can preserve non-deterministic nature of inputs and performances as well as handle incomplete or partial knowledge of experts or citizens, i.e. "I don't know" type opinion. The "degrees of belief" measures allow experts to express their strength of opinions in the conservative and optimistic terms. Such operation is not possible by the probability-based approach. • Dempster-Shafer theory can avoid the scalability issue encountered in Bayesian reasoning. It can also measure uncertainty in the reasoning chains, and identify information needed for improving the reasoning process. • Use of Dempster's rule of combination, instead of the average operator in probability theory, to merge expert opinions about inputs or relations is a better way for combining conflicting and incomplete opinions. In the dissertation, the Belief Reasoning method is applied in real-world Alternatives Analysis of a transit investment project. The results show its potential to analyze and evaluate the alternatives and to provide reasons for recommending a preferred alternative and to measure the uncertainty in the reasoning process. In spite of some shortcomings, discussed in the dissertation, the Belief Reasoning method is an effective method for transportation planning compared with the existing methods. It provides means for the planners and citizens to present their own reasons and allows review and analysis of reasoning and judgments of all participating stakeholders. The proposed method can promote focused discourse among different groups of stakeholders, and enriches the quality of the planning process. / Ph. D.

The purpose of this dissertation is to develop and demonstrate methods appropriate for the quantification and propagation of uncertainty in large, high-consequence engineering projects. The term "rigorous uncertainty quantification" refers to methods equal to the proposed task. The motivating practical example is uncertainty in a Mars atmosphere model due to the incompletely characterized presence of dust. The contributions made in this dissertation, though primarily mathematical and philosophical, are driven by the immediate needs of engineers applying uncertainty quantification in the field. Arguments are provided to explain how the practical needs of engineering projects like Mars lander missions motivate the use of the objective probability bounds approach, as opposed to the subjectivist theories which dominate uncertainty quantification in many research communities. An expanded formalism for Dempster-Shafer structures is introduced, allowing for the representation of continuous random variables and fuzzy variables as Dempster-Shafer structures. Then, the correctness and incorrectness of probability bounds analysis and the Cartesian product propagation method for Dempster-Shafer structures under certain dependency conditions are proven. It is also conclusively demonstrated that there exist some probability bounds problems in which the best-possible bounds on probability can not be represented using Dempster-Shafer structures. Nevertheless, Dempster-Shafer theory is shown to provide a useful mathematical framework for a wide range of probability bounds problems. The dissertation concludes with the application of these new methods to the problem of propagating uncertainty from the dust parameters in a Mars atmosphere model to uncertainty in that model's prediction of atmospheric density. A thirty-day simulation of the weather at Holden Crater on Mars is conducted using a meso-scale atmosphere model, MRAMS. Although this analysis only addresses one component of Mars atmosphere uncertainty, it demonstrates the applicability of probability bounds methods in practical engineering work. More importantly, the Mars atmosphere uncertainty analysis provides a framework in which to conclusively establish the practical importance of epistemology in rigorous uncertainty quantification. / Ph. D.

Management of data imprecision has become increasingly important, especially with the advance of technology enabling applications to collect and store huge amount data from multiple sources. Data collected in such applications involve a large number of variables and various types of data imperfections. These data, when used in knowledge discovery applications, require the following: 1) computationally efficient algorithms that works faster with limited resources, 2) an effective methodology for modeling data imperfections and 3) procedures for enabling knowledge discovery and quantifying and propagating partial or incomplete knowledge throughout the decision-making process. Bayesian Networks (BNs) provide a convenient framework for modeling these applications probabilistically enabling a compact representation of the joint probability distribution involving large numbers of variables. BNs also form the foundation for a number of computationally efficient algorithms for making inferences. The underlying probabilistic approach however is not sufficiently capable of handling the wider range of data imperfections that may appear in many new applications (e.g., medical data). Dempster-Shafer theory on the other hand provides a strong framework for modeling a broader range of data imperfections. However, it must overcome the challenge of a potentially enormous computational burden. In this dissertation, we introduce the joint Dirichlet BoE, a certain mass assignment in the DS theoretic framework, that simplifies the computational complexity while enabling one to model many common types of data imperfections. We first use this Dirichlet BoE model to enhance the performance of the EM algorithm used in learning BN parameters from data with missing values. To form a framework of reasoning with the Dirichlet BoE, the DS theoretic notions of conditionals, independence and conditional independence are revisited. These notions are then used to develop the DS-BN, a BN-like graphical model in the DS theoretic framework, that enables a compact representation of the joint Dirichlet BoE. We also show how one may use the DS-BN in different types of reasoning tasks. A local message passing scheme is developed for efficient propagation of evidence in the DS-BN. We also extend the use of the joint Dirichlet BoE to Markov models and hidden Markov models to address the uncertainty arising due to inadequate training data. Finally, we present the results of various experiments carried out on synthetically generated data sets as well as data sets from medical applications.