PhD thesis abstracts

PhD thesis abstracts

• Andre Miede
• Beatriz Soret
• Lin Lin
• Peter Knees
• Vineeth N Balasubramanian

Andre Miede

Cross-Organizational Service Security - Attack Modeling and Evaluation of Selected Countermeasures

Challenging market dynamics and the rise of complex value networks require organizations to adjust their processes rapidly in order to stay competitive. Because many organizational processes are directly supported or even enabled by Information Technology (IT), a process is only as flexible as its underlying technological representation. The Service-oriented Architecture paradigm (SOA) offers means on both a technological and organizational level for the flexible integration of internal and external IT systems. Thus, services are used to assemble processes through service compositions, as well as across enterprise boundaries. Such cross-organizational service-based workflows lead to a global SOA which is often referred to as the "Internet of Services".

Just as any economic system requires security in order to function and to be accepted by its participants, the security of the involved IT systems, exchanged messages, and communication channels used has to be ensured for cross-organizational service-based collaboration. Achieving and guaranteeing basic IT security goals such as confidentiality, authentication, authorization, non-repudiation, integrity, availability, and anonymity is a necessity in this context and an active topic, both in research and industry.

The main tenor of current SOA security research is that conventional security measures are not effective enough in the SOA context. Furthermore, just equalizing SOA security with Web service security reduces SOA security requirements to Web service security standards and their configuration, which is an incomplete view.

This thesis makes several contributions regarding the security of service-based systems: First, it is shown how a model of cross-organizational SOA concepts can be used for analyzing SOA elements regarding their impact on security. This is done by applying core IT security concepts, such as threats, vulnerabilities, etc., to the general elements of a cross-organizational SOA, such as loose coupling, composability, etc.

Second, an analysis of attacks in the Internet of Services is performed by proposing an attack taxonomy for service-based systems and by modeling selected examples of service-specific attack classes. This goes beyond the current state-of-the-art regarding SOA attacks by taking into account more service-specific and business-oriented threats. The modeling of these attacks builds on a self-developed generic metamodel, that brings together the most important concepts of IT security and their relationships. It is shown, how assets, threats, vulnerabilities, risks, security goals, etc. relate to each other at the core of this metamodel and what the basic structure of countermeasures is.

Third, an attack scenario of communication analysis that threatens relationship anonymity in the Internet of Services is further investigated, due to its system-inherent implications. With a particular focus on service compositions, a simulation-based evaluation of different attack models and scenarios offers insights regarding the anonymity of cross-organizational collaboration. Furthermore, the impact of using standard anonymity mechanisms on selected Quality of Service parameters is evaluated for Web services in real networks. The obtained results aim at identifying the limits of anonymity in the Internet of Services and at quantifying side-effects of using state-of-the-art countermeasures.

Beatriz Soret

Analysis of QoS parameters in fading channels based on the effective bandwidth theory

Providing Quality of Service (QoS) guarantees is an important challenge in the design of next generations of wireless networks. In particular, real-time services involving stringent delay constraints are expected to be increasingly popular among users of mobile equipments. In Rayleigh channels, the delay requirement is usually expressed in terms of a probabilistic delay constraint composed by two terms: the target delay and the probability of exceeding the target delay.

In this thesis, a variable-rate multiuser and multichannel system using adaptive modulation is addressed. Specifically, the tradeoff among information source, fading channel and delay is analyzed, based on the effective bandwidth theory. Within the effective bandwidth framework, expressions of the channel effective bandwidth function (also known as effective capacity) are obtained on the channel side. Several scenarios are addressed: uncorrelated and time-correlated flat Rayleigh channels and an OFDM system under a frequency-selective Rayleigh channels. The procedure to obtain these functions is generic and could be applied to other channel models and scenarios.

The effective bandwidth theory makes feasible the analysis of the distribution tail of the delay. The percentile of the delay and the maximum information rate that can be transmitted over the channel under a target BER and a probabilistic delay constraint are evaluated. The delay suffered by certain information flow depends not only on the transmission rate but also on the distribution and self-correlation of the information process. Even in wired systems (constant rate channels) different distributions of the information process having the same average rate will cause different delays. Indeed, the better conditions for the delay are obtained when the incoming user traffic is constant. For any other source process, the delay performance degrades. Besides, the correlation of the channel process, in time or in frequency, has also a negative impact on the delay.

In the last part of the thesis, multiplexing of users over multiple shared fading channel is addressed. A new element comes up in this case: the scheduling algorithm. We calculate the maximum rate that each user can transmit by fulfilling a target BER and its own delay constraint, and under a given scheduling discipline. The analysis is done first in a single channel link and later on generalized to multiple shared channels employing OFDMA as multiplexing mechanism. Now it is not only the delay constraint and the channel and source process that influence the source rate, but also the discipline that rules the system. Three representative multiplexing algorithms are analyzed: Round Robin, Best Channel and Proportional Fair. The results make possible the comparison of the algorithms in terms of throughput, delay and fairness.

In summary, this thesis shows the high sensitivity of the delay to the burstiness of the traffic, to the time or frequency correlation of the channel and to the scheduling discipline. The proposed procedure is generic and can be extended to other disciplines and traffic and channel models. Nevertheless, the effective bandwidth function of the source and the channel process cannot always be explicitly evaluated. For such cases, a semi-analytical strategy is also proposed.

Lin Lin

Multimedia Data Mining and Retrieval for Multimedia Databases Using Associations and Correlations

With the explosion in the complexity and amount of pervasive multimedia data, there are high demands of multimedia services and applications in various areas for people to easily access and distribute multimedia data. Facing with abundance multimedia resources but inefficient and rather old-fashioned keyword-based information retrieval approaches, a content-based multimedia information retrieval (CBMIR) system is required to (i) reduce the dimension space for storage saving and computation reduction; (ii) advance multimedia learning methods to accurately identify target semantics for bridging the semantics between low-level/mid-level features and high-level semantics; and (iii) effectively search media content for dynamical media delivery and enable the extensive applications to be media-type driven.

This research mainly focuses on multimedia data mining and retrieval system for multimedia databases by addressing some main challenges, such as data imbalance, data quality, semantic gap, user subjectivity and searching issues. Therefore, a novel CBMIR system is proposed in this dissertation. The proposed system utilizes both association rule mining (ARM) technique and multiple correspondence analysis (MCA) technique by taking into account both pattern discovery and statistical analysis. First, media content is represented by the global and local low-level and mid-level features and stored in the multimedia database. Second, a data filtering component is proposed in the system to improve the data quality and reduce the data imbalance. To be specific, the proposed filtering step is able to vertically select features and horizontally prune instances in multimedia databases. Third, a new learning and classification method mining weighted association rules is proposed in the retrieval system. The MCA-based correlation is used to generate and select the weighted N-feature-value pair rules, where the N varies from 1 to many. Forth, a ranking method independent of classifiers is proposed in the system to sort the retrieved results and put the most interesting ones on the top of the browsing list. Finally, a user interface is implemented in CBMIR system that allows the user to choose his/her interested concept, searches media based on the target concept, ranks the retrieved segments using the proposed ranking algorithm, and then displays the top-ranked segments to the user.

The system is experimented with various high-level semantics from TRECVID benchmark data sets. TRECVID sound and vision data is a large data set, includes various types of videos, and has very rich semantics. Overall, the proposed system achieves promising results in comparison with the other well-known methods. Moreover, experiments that compare each component with some other famous algorithms are conducted. The experimental results show that all proposed components improve the functionalities of the CBMIR system, and the proposed system reaches effectiveness, robustness and efficiency for a high-dimensional multimedia database.

Peter Knees

Text-Based Description of Music for Indexing, Retrieval, and Browsing

The aim of this PhD thesis is to develop automatic methods that extract textual descriptions from the Web that can be associated with music pieces. Deriving descriptors for music permits to index large repositories with a diverse set of labels and allows for retrieving pieces and browsing collections. The techniques presented make use of common Web search engines to find related text content on the Web. From this content, descriptors are extracted that may serve as

• labels that facilitate orientation within browsing interfaces to music collections, especially in a three-dimensional browsing interface presented,

ind

• exing terms, used as features in music retrieval systems that can be queried using descriptive free-form text as input, and

featur

• es in adaptive retrieval systems that aim at providing more user-targeted recommendations based on the user's searching behaviour for exploration of music collections.

In the context of this thesis, different extraction, indexing, and retrieval strategies are elaborated and evaluated. Furthermore, the potential of complementing Web-based retrieval with acoustic similarity extracted from the audio signal, as well as complementing audio-similarity-based browsing approaches with Web-based descriptors is investigated and demonstrated in prototype applications.

Vineeth N Balasubramanian

Conformal Predictions in Multimedia Pattern Recognition

The field of multimedia pattern recognition is on a fundamental quest to design intelligent systems that can learn and behave the way humans do. One important aspect of human intelligence that has so far not been given sufficient attention in these fields is the capability of humans to hedge decisions. Humans can express when they are certain about a decision they have made, and when they are not. Unfortunately, machine learning techniques today are not yet fully equipped to be trusted with this critical task. This work seeks to address this fundamental knowledge gap. Existing approaches that provide a measure of confidence of a learning algorithm on a prediction such as those based on the Bayesian theory or the Probably Approximately Correct learning theory require strong assumptions or often produce results that are not practical or reliable. However, the recently developed Conformal Predictions (CP) framework - which is based on the principles of hypothesis testing, transductive inference and algorithmic randomness - provides a game-theoretic approach to the estimation of confidence with several desirable properties such as online calibration and generalizability to all classification and regression methods.

This dissertation builds on the theory of Conformal Predictions to compute reliable confidence measures that aid decision-making in real-world multimedia problems. The theory behind the CP framework guarantees that the confidence values obtained using this transductive inference framework manifest as the actual error frequencies in the online setting, i.e. they are well-calibrated. Further, this framework can be used with any classifier, meta-classifier or regressor (such as Support Vector Machines, k-Nearest Neighbors, Adaboost, ridge regression, etc). The key contributions of this dissertation (outlined below) are validated on four problems from the domains of healthcare and assistive technologies: two classification-based applications (risk prediction in cardiac decision support and multimodal person recognition), and two regression-based applications (head pose estimation and saliency prediction in radiological images). The cost of errors in decision-making is often high in these application domains, and hence these problems are selected to validate the contributions. The key contributions of this work are summarized below:

(1) Efficiency Maximization in Conformal Predictors: The CP framework has two important properties that define its utility: validity and efficiency. Validity refers to controlling the frequency of errors within a pre-specified error threshold. Also, since the framework outputs a set of possible predictions as the result, it is essential that the prediction sets are as small as possible. This property is called efficiency. Evidently, an ideal implementation of the framework would ensure that the algorithm provides high efficiency along with validity. However, this is not a straightforward task, and depends on the learning algorithm (classification or regression, as the case may be) as well as the non-conformity measure chosen in a given context. In this work, a novel framework to learn a kernel (or distance metric) that will maximize the efficiency in a given context has been proposed and validated on different risk-sensitive applications.

(2) Conformal Predictions for Information Fusion: The CP framework ensures the calibration property in the estimation of confidence in pattern recognition. Most of the existing work in this context has been carried out using single classification systems or ensemble classifiers (such as boosting). However, there been a recent growth in the use of multimodal fusion algorithms and multiple classifier systems. A study of statistical approaches to combine p-values from multiple classifiers and regressors has been performed, which revealed the usefulness of quantile combination methods for calibrated confidence values in information fusion contexts.

(3) Online Active Learning using Conformal Predictors: As newer data are encountered, it becomes essential to select appropriate data instances for labeling and updating the classifier to facilitate a continuously learning system. Using the p-values computed by the CP framework, a novel online active learning approach has been proposed and validated. This active learning method can also be extended to an information fusion setting, where there are multiple information sources or multiple modalities.

The results obtained in this work demonstrate promise and potential in using these contributions to provide reliable measures of confidence in multimedia pattern recognition problems in real-world settings.