Διδακτορικές διατριβές
Μόνιμο URI για αυτήν τη συλλογήhttps://pyxida.aueb.gr/handle/123456789/5
Περιήγηση
Πρόσφατες Υποβολές
Τεκμήριο Μέθοδοι βελτίωσης αποτελεσμάτων αναζήτησης στον παγκόσμιο ιστόΚρητικόπουλος, Απόστολος; Οικονομικό Πανεπιστήμιο Αθηνών, Τμήμα Πληροφορικής; Σιδέρη, ΜάρθαΗ ποιότητα μιας μηχανής αναζήτησης εξαρτάται κυρίως από τα 10 πρώτα αποτελέσματα που παρουσιάζονται ως απάντηση σε ένα ερώτημα του χρήστη. Η σειρά παρουσίασης των αποτελεσμάτων έχει άμεση σχέση με τον αλγόριθμο βαθμολόγησης (rating) ιστοσελίδων της μηχανής αναζήτησης.Στην διατριβή αυτή προτείνονται νέες μέθοδοι βαθμολόγησης ιστοσελίδων που αποσκοπούν στη βελτίωση της ποιότητας των αποτελεσμάτων των μηχανών αναζήτησης, και στην εξατομίκευση των αναζητήσεων στον Παγκόσμιο Ιστό (World Wide Web). Προτείνεται ένα μέτρο αξιολόγησης των αποτελεσμάτων της αναζήτησης που δεν απαιτεί ρητή πληροφορία από τους χρήστες, και αξιοποιεί τα υποκρυπτόμενα χαρακτηριστικά της συμπεριφοράς τους. Τέλος παρουσιάζεται μία κατανεμημένη εφαρμογή περιήγησης του Web η οποία επιλεκτικά συλλέγει ιστοσελίδες, καθώς και η υλοποίηση δύο μηχανώναναζήτησης. Συγκεκριμένα, προτείνονται τρεις αλγόριθμοι: ο Compass Filter, ο WordRank και ο BlogRank. Ο Compass Filter είναι αλγόριθμος που εξατομικεύει τα αποτελέσματα της αναζήτησης, και δίνει προτεραιότητα στις ιστοσελίδες των διαδικτυακών κοινοτήτων που έχει ήδη επισκεφτεί ο χρήστης στο παρελθόν. Ο WordRank βαθμολογεί τις ιστοσελίδες δίνοντας προτεραιότητα σε αυτές που συνδέονται μεταξύ τους, και έχουν όμοιο περιεχόμενο. Ο BlogRank αξιοποιεί τα ιδιαίτερα χαρακτηριστικά των blogs (όπως συντάκτες, θέματα, ειδήσεις, ημερομηνίες δημοσίευσης) και τα βαθμολογεί βάσει των κοινών ιδιοτήτων τους. Αποδεικνύεται πειραματικά ότι οι τρεις αυτοί αλγόριθμοι βελτιώνουν σημαντικά τα αποτελέσματα που παρουσιάζονται στους χρήστες. Στα πειράματα παρουσιάζεται ο Success Index, ένα νέο μέτρο για την αξιολόγηση της ποιότητας των αποτελεσμάτων μιας μηχανής αναζήτησης. Ο Success Index χρησιμοποιεί την υποκρυπτόμενη κριτική (implicit feedback) των χρηστών για να απεικονίσει την ποιότητα των αποτελεσμάτων. Θεωρούμε ότι ο χρήστης εκφράζει την κριτική του στααποτελέσματα μίας αναζήτησης, με τις ιστοσελίδες που επιλέγει να δει (ποιες σελίδες επιλέγει, και με ποια σειρά). ∆είχνουμε πειραματικά ότι ο βαθμός που αποδίδει ο Success Index είναι παρόμοιος με αυτόν μιας ρητής (explicit) αξιολόγησης. Για τον έλεγχο των προτεινόμενων αλγορίθμων και του μέτρου αξιολόγησης αναπτύχθηκαν δύο μηχανές αναζήτησης, και ένας κατανεμημένος παραμετρικός περιηγητής ιστού (web crawler).Τεκμήριο Resource allocation, content recommendations and online learning mechanisms for mobile edge computing(05/23/2022) Χατζηελευθερίου, Λίβια-Έλενα; Athens University of Economics and Business, Department of Informatics; Polyzos, George; Toumpis, Stavros; Iosifidis, George; Dimakis, Antonios; Siris, Vasileios; Stamoulis, Georgios; Koutsopoulos, IordanisThe Mobile Edge Computing (MEC) paradigm brings computing and cache capacity resources in the proximity of users. It gives rise to a new ecosystem of services, such as Augmented Reality (AR) ones, while reducing the latency that is experienced by users and lowering network service costs. The main challenges that MECfaces are related to the scarcity of resources at the network edge, the unpredictability of important system parameters, such as traffic, content and computation demand, and the ultra-low latency requirements that must be satisfied.In this thesis we deal with the challenges above, towards the optimization of two MEC goals: content delivery and real-time analytics at the edge of the network. We present resource allocation mechanisms and methods that automate the resource allocation, for fifth-generation (5G), Beyond-5G (B5G) and sixth-generation (6G) communication systems, accounting for edge resources such as caches, computational resources of mobiledevices and edge servers, bandwidth and energy. We tackle both offline and Online Learning (OL) instances of optimization problems that span content recommendationsand caching, user association and allocation of computing resources. We use a variety of mathematical tools to solve these problems, such as combinatorial optimization, convex optimization and Online Convex Optimization (OCO), a special case of OL. We analyse and we exploit the structural properties of the formulated optimization problems, either by designing algorithms ex novo, or by adapting existing techniques to our settings. We provide cost-efficient, fast and elegant solutions with provable performance guarantees, for a variety of important problems that arise within the MEC context. Overall, this Ph.D. thesis tackles a set of important optimization problems that arise in the context of edge computing and networking. We present novel problem formulations and algorithms that lead to solutions with provable performance guarantees, bringing the Mobile Edge Computing (MEC) paradigm a step closer to its practical realization.Τεκμήριο Mechanisms and policies for resource allocation in clouds and next generation networks(02/24/2022) Darzanos, George; Νταρζάνος, Γεώργιος; Athens University of Economics and Business, Department of Informatics; Koutsopoulos, Iordanis; Courcoubetis, Costas A.; Polyzos, George C.; Siris, Vasilios A.; Dimakis, Antonios; Denazis, Spyros; Stamoulis, George D.Το οικοσύστημα του Διαδικτύου σήμερα αποτελείται από ένα μεγάλο εύρος φορέων όπως οι Πάροχοι Δικτύων Κινητής Τηλεφωνίας (Mobile Network Operators - MNOs), οι Πάροχοι Υπηρεσιών Υπολογιστικής Νέφους (Cloud Service Providers - CSPs), τα Δίκτυα Διανομής Περιεχομένου (Content Delivery Networks - CDNs), οι Πάροχοι Περιεχομένου (Content Providers - CPs), οι Πάροχοι Υπηρεσιών Over-the-Top (OTT) κ.λ.π.. Οι φορείς αυτοί προσφέρουν αμοιβαίως συμπληρωματικές λειτουργίες και αλληλεπιδρούν μεταξύ τους για την από κοινού παροχή υπηρεσιών προστιθέμενης αξίας σε επιχειρήσεις και τελικούς χρήστες. Ωστόσο, αυτοί οι φορείς είναι ορθολογικές (rational) οντότητες που αναζητούν κέρδη. Συνεπώς, αν και συνεργάζονται λειτουργικά, ανταγωνίζονται μεταξύ τους για τη διανομή των εσοδών από τους πελάτες των προσφερόμενων υπηρεσιών. Στην παρούσα διδακτορική διατριβή, προτείνουμε και αξιολογούμε υποδείγματα και μηχανισμούς που διευκολύνουν την αλληλεπίδραση αυτών των φορέων, καθώς και μηχανισμούς και πολιτικές που επιτυγχάνουν αποτελεσματική κατανομή πόρων και εσόδων σε συνεργατικά και σε ανταγωνιστικά οικοσυστήματα, καθώς και σε ενδιάμεσες περιπτώσεις συνεργασίας. Ειδικότερα, μελετάμε: (i) τη δημιουργία ομοσπονδιών από πολλαπλούς CSPs, εστιάζοντας στην παροχή υπηρεσιών υπολογιστικής νέφους που αξιοποιούν τους συλλογικά διαθέσιμους υπολογιστικούς πόρους, (ii) τις αλληλεπιδράσεις πολλαπλών MNOs κατά τη δημιουργία τεμαχίων δικτύου (network slices) τα οποία πρέπει να διασχίζουν πολλούς MNOs προκειμένου να καταστεί δυνατή η εξυπηρέτηση των αντίστοιχων από-άκρο-σε-άκρο εφαρμογών, και (iii) την από κοινού εξέταση της ανάθεσης χρηστών και της τοποθέτησης περιεχομένου σε Σταθμούς Βάσης Μικρού Κελιού (Small-cell Base Station - SBS), όπου ένας MNO και ένας CP συνεργάζονται για την αποτελεσματική διαχείριση των διαθέσιμων πόρων του ασύρματου δικτύου και της χωρητικότητας αποθήκευσης περιεχομένου (caches).Τεκμήριο Spectral analysis and preconditioned iterative solvers for large structured linear systems(03/16/2022) Barakitis, Nikos; Μπαρακίτης, Νικόλαος; Athens University of Economics and Business, Department of Informatics; Serra-Capizzano, Stefano; Vrahatis, Michael N.; Gallopoulos, Efstratios; Yannacopoulos, Athanasios; Toumpis, Stavros; Triantafyllou, Dimitrios; Vassalos, ParisIn this thesis, the numerical solution of three different classes of problems have been studied. Specifically, new techniques have been proposed and their theoretical analysis has been performed, accompanied by a wide set of numerical experiments, for investigating further and comparing the effectiveness and performance of the presented approach. The first two belong to the research area of numerical linear algebra and concern the spectral analysis and preconditioning for Krylov subspace methods of the coefficient matrix of large structured linear systems. The third concerns a problem from the area of financial computing namely the pricing of an American put option.In the first set of problems, the asymptotic spectra of large matrices coming from the summarization of Toeplitz structure functions was studied. The spectral asymptotic behavior of this matrix sequences was provided analytically. Taking advantage of this analysis circulant preconditioners were proposed, and the eigenvalue distribution of the preconditioned matrix sequences was given. All theoretical results were numerically confirmed. The second problem that was studied concerns the theoretical and numerical exploration of proper preconditioners based on the spectral symbols of the coefficient matrix arising from the discretization of Fractional order Differential Equations problems. Beside the theoretical study, a comparison between the already propose in the literature techniques was conducted. The numerical experiments show that in the one dimensional case, the proposed preconditioners perform similarly with the best known techniques for this problem. However, in the challenging and more interesting for the applications multivariate setting the proposed preconditioners show their superiority against all the competitors.For the pricing of an American put option an iterative algorithm based on the theory of dynamic programming was proposed. Taking advantage of the already known characteristics of the optimal value function it was proved theoretically and numerically confirmed that the proposed algorithm obtains monotonically increasing value functions and converges to the optimal one.Τεκμήριο Quality of musicians' experience in network music performance(03/16/2022) Tsioutas, Konstantinos; Τσιούτας, Κωνσταντίνος; Athens University of Economics and Business, Department of Informatics; Polyzos, George C.; Siris, Vasilios A.; Apostolopoulos, Theodoros; Doumanis, Ioannis; Kalogeraki, Vana; Floros, Andreas; Xylomenos, GeorgeThe increased use of tele-presence and tele-conferencing facilities, whether due to the need to isolate during a pandemic, or due to the desire to avoid costly and time consuming travel, prompted a renewed interest in Network Music Performance (NMP), where musicians collaborate remotely over the Internet in real time. Although the Internet has made dramatic leaps in capacity since the first NMP systems were created in the 20th century, the delays involved when communicating over the Internet, whether due to the physical distance between the endpoints, or due to the unpredictable nature of network traffic, are an important hindrance to the widespread use of NMP applications.The main question that this thesis attempts to answer is how much delay humans are able to tolerate for NMP to be acceptable. To achieve this goal, we first identify the factors influencing the Quality of Musicians' Experience (QoME) during NMP. Out of these factors, we single out audio delay, which makes or breaks a performance. We also consider audio quality, as it may be reduced to save bandwidth, without resorting to delay-inducing audio compression. A review of the literature shows that past work on evaluating the human tolerance to delay during NMP either employs a scenario where music is not performed, that is, synchronization of hand claps, or involves a very small number of experiments, thus having low statistical significance.Before embarking on a large scale study of NMP with actual musical performances, we first performed two exploratory studies. The first study tested our experimental setup, including the software and hardware employed, so as to ensure that the testing environment was acceptable to musicians and that we could gather accurate data without interruptions. The second study tested our assessment method, which consisted of questionnaires answered by each participant at the end of every performance, with a small number of musicians. Based on these studies, we then designed and carried out the largest NMP study to date with actual musicians performing real musical pieces. In this study, we varied either audio delay or audio quality in a systematic manner, gathering up answers to a fine-tuned questionnaire for QoME assessment. This subjective evaluation revealed that after crossing a quality threshold, further increasing audio quality had no discernible effects to QoME, indicating that when bandwidth is limited, we can sacrifice (up to a point) audio quality to reduce the required bitrate, without resorting to compression. On the other hand, we found that varying delay did have a statistically significant effect to QoME. More importantly though, our results indicate that the delay threshold up to which NMP is feasible is closer to 40~ms, rather than the 25-30~ms previously considered acceptable. Having recorded audio and video from all sessions, we complemented this subjective study with three additional evaluation methods, making our work the first multimodal study of the QoME for NMP. First, we performed tempo analysis on the recorded audio, to assess the highest delay at which the musicians could maintain a steady tempo; the results from this study confirmed that delays of up to 40~ms are acceptable for NMP, as indicated by the subjective study. Second, we analyzed the audio features of the recordings, finding that delay had a larger impact on percussive instruments and musicians performing rhythm parts; this result confirmed similar results from a previous, but much smaller study. Third, we analyzed the video recordings in order to detect the emotions felt by the musicians using machine learning methods, finding that as audio delay or audio quality was varied there were clear disruptions in the emotions of the musicians; while these results are intriguing, they were not clear enough to substitute the subjective analysis.Τεκμήριο Developing resilience and cyber-physical protection capabilities for critical aviation infrastructures(06/14/2021) Lykou, Georgia; Λύκου, Γεωργία; Athens University of Economics and Business, Department of Informatics; Apostolopoulos, Theodoros; Stamatiou, Υannis; Μαυρίδης, Ιωάννης; Μάγκος, Εμμανουήλ; Κοτζανικολάου, Παναγιώτης; Στεργιόπουλος, Γεώργιος; Gritzalis, DimitriosTransport sector is a critical infrastructure that greatly supports the smooth functioning of society's welfare and viability of economies worldwide. Disruptions to transportation systems can cause large economic impacts or even human losses, so they should be adequately protected from physical and cyber-physical threats.We focused our research on aviation sector, which is the safest transport mode, however the most interdepended one in terms of information and communication technologies applied. Cyber-attacks are increasing in quantity and persistence, so the consequences of a successful malicious cyber-attack to civil aviation operations could be severe nowadays. Aiming to enhance operational practices and develop robust cybersecurity governance in smart airports, we have presented a systematic and comprehensive analysis of unlawful attacks towards smart airports, by implementing cybersecurity best practices and resilience measures.Our research examined cyber security challenges and interoperability in Air Traffic Management systems and proposed an extended threat model for analysing possible targets and risks involved. We analysed cyber resilience aspects in the aviation context and the need for holistic strategy of defence, prevention, and response. Furthermore, as the fastest growing segment of aviation, Unmanned Aerial Systems(UAS) continue to increase in technical complexity, and capabilities. However, UAS pose significant challenges in terms of safety, security, and privacy. An increasing phenomenon, nowadays, is drone-related incidents near airport facilities, which are expected to proliferate in frequency and severity, as drones become larger and more powerful. Critical infrastructures need to be protected from such aerial attacks, through effective counteracting technologies, risk management, and resilience plans. In this dissertation, we have explored how counter drone technologies can prevent, detect, identify, and mitigate rogue drones. We have analysed realistic attack scenarios of malicious drones’ attacks and proposed an effective C-UAS protection plan for each case. We have also discussed the applicability limitations of C-UAS in the aviation context and proposed a resilience action plan for airports stakeholders for defending airborne threats from misused drones.The integration of our research in the aviation sector, focused on air transport networks and introduced a risk-based method to analyse interdependencies and congestions in the aviation network. The proposed methodology and software tool can assess delay incidents in airports, produce weighted risk dependency graphs, presenting how a delay that occurred in one airport may affect the operational efficiency of other interconnected airports. The tool can also detect the most critical airports and congested connections, while it can indicate the n-order dependency chains, which should be avoided by airline flight planners, to reduce delay impacts in the aviation network.Τεκμήριο Deep neural networks for information mining from legal texts(04/19/2021) Chalkidis, Ilias; Χαλκίδης, Ηλίας; Athens University of Economics and Business, Department of Informatics; Aletras, Nikolaos; Vazirgiannis, Michalis; Vassalos, Vasilios; Koubarakis, Manolis; Konstas, Ioannis; Paliouras, Georgios; Androutsopoulos, IonLegal text processing (Ashley, 2017) is a growing research area where Natural Language Processing (NLP) techniques are applied in the legal domain. There are several applications such as legal text segmentation (Mencia, 2009; Hasan et al., 2008), legal topic classification (Mencia and Fürnkranzand, 2007; Nallapati and Manning, 2008), legal judgment prediction and analysis (Wang et al., 2012; Aletras et al., 2016), legal information extraction (Kiyavitskaya et al., 2008; Dozier et al., 2010; Asooja et al., 2015), and legal question answering (Kim et al., 2015b, 2016b). These applications and relevant NLP techniques arise from three main sub-domains, i.e, legislation, court cases, and legal agreements (contracts). In all three sub-domains, documents are much longer than in most other modern NLP applications. They also have different characteristics concerning the use of language, the writing style, and their structuring, compared to non-legal text. Given the rapid growth of deep learning technologies (Goodfellow et al., 2016; Goldberg, 2017), the goal of this thesis is to explore and advance deep learning methods for legal tasks, such as contract element and obligation extraction, legal judgment prediction, legal topic classification, and information retrieval, that have already been discussed in the literature (but not in the context of deep learning) or that were first addressed during the work of this thesis. In this direction, we aim to answer two main research questions: First and foremost on the adaptability of neural methods that have been proposed for related NLP tasks in other domains and how they are affected by legal language, writing, and structure; and second on providing explanations of neural models’ decisions (predictions). Considering the first research question we find and highlight several cases, where either legal language affects a model’s performance or suitable modeling is needed to imitate the document structure. To this end, we pre-train and use in-domain word representations and neural language models, while we also propose new methods with state-of-the-art performance. With respect to model explainability, we initially experiment with saliency (attention) heat-maps and highlight their limitations as a means for the explanation of the model’s decisions, especially in the most challenging task of legal judgment prediction, where it is most important. To overcome these limitations we further study rationale extraction techniques as a prominent methodology towards model explainability.In lack of publicly available annotated datasets in order to experiment with deep learning methods, we curate and publish five datasets for various legal tasks (contract element extraction, legal topic classification, legal judgment prediction and rationale extraction, and legal information retrieval), while we also publish legal word embeddings and a legal pre-trained language model to assist legal text processing research and development. We consider our work, a first, fundamental, step among other recent efforts, towards improving legal natural language understanding using state-of-the-art deep learning techniques, which further promotes the adaptation of new technologies and sheds light on the emerging field of legal text processing.Τεκμήριο Organisational energy conservation behaviour & the effect of motivational information systems and gamification on employees’ energy savingΚωτσόπουλος, Δημοσθένης; Kotsopoulos, Dimosthenis; Athnes University of Economics and Business, Department of Informatics; Πουλούδη, Αθανασία; Ιορδάνογλου, Δήμητρα; Κουρουθανάσης, Παναγιώτης; Πραματάρη, Αικατερίνη; Δουκίδης, Γεώργιος; Λεκάκος, ΓεώργιοςEnergy consumption is one of the widely recognised important factors that should be addressed, in reducing CO2 emissions towards protecting the environment and addressing climate change. However, albeit the documented effect of public buildings on energy consumption, only a limited body of research focuses on one of the most important factors that could limit it: employees’ energy consumption behaviour. Moreover, motivational information systems and gamification, the use of game elements in non-game contexts, when appropriately utilised, could lead to positive employees’ energy behaviour change. This research focuses on addressing two research questions: Q1: What drives employees to save energy at work, and Q2: How effective is IoT-enabled gamified feedback in producing energy behaviour change and energy savings at work? Aiming to contribute to this path of research, and adopting a mixed-methods research approach, we conducted a series of five (5) studies, to investigate employees’ energy-saving behaviour at work and explore the effect of motivational Information Systems featuring personalized gamified feedback enabled by Internet-of-Things (IoT) technologies, towards conserving energy at work.Τεκμήριο Aspect Based Sentiment AnalysisPavlopoulos, Ioannis; Athens School of Economics and Business, Department of Informatics; Androutsopoulos, IonAspect Based Sentiment Analysis (ABSA) systems receive as input a set of texts (e.g.,product reviews or messages from social media) discussing a particular entity (e.g., anew model of a mobile phone). The systems attempt to detect the main (e.g., the mostfrequently discussed) aspects (features) of the entity (e.g., ‘battery’, ‘screen’) and toestimate the average sentiment of the texts per aspect (e.g., how positive or negativethe opinions are on average for each aspect). Although several ABSA systems have been proposed, mostly research prototypes, there is no established task decompositionfor ABSA, nore are there any established evaluation measures for the subtasks ABSA systems are required to perform. This thesis, proposes a new task decomposition for ABSA, which contains three main subtasks: aspect term extraction, aspect term aggregation, and aspect term polarity estimation. The first subtask detects single- and multi-word terms naming aspects of the entity being discussed (e.g., ‘battery’, ‘hard disc’), called aspect terms. The second subtask aggregates (clusters) similar aspect terms (e.g., ‘price’ and ‘cost’, but maybe also ‘design’ and ‘color’), depending on user preferences and other restrictions (e.g., the size of the screen where the results of the ABSA system will be shown). The third subtask estimates the average sentiment per aspect term or cluster of aspect terms. For each one of the above mentioned subtasks, benchmark datasets for different kinds of entities (e.g., laptops, restaurants) were constructed during the work of this thesis. New evaluation measures are introduced for each subtask, arguing that they are more appropriate than previous evaluation measures. For each subtask, the thesis also proposes new methods (or improvements over previous methods), showing experimentally on the constructed benchmark datasets that the new methods (or the improved versions) are better or at least comparable to state of the art ones.Τεκμήριο Ικανές συνθήκες για την ύπαρξη υπογραφημάτων με καθορισμένη δομή σε ειδικές κατηγορίες γραφημάτωνΦουρτουνέλλη, Όλγα; Athens School of Economics and Business, Department of Informatics; Κατερίνης, ΠαναγιώτηςThe purpose of this thesis is to establish sufficient conditions for the existence of factors with prescribed properties in special classes of graphs; mainly in terms of the graph invariants of connectivity and minimum degree. The thesis starts by examining the relation between the minimum degree of a claw-free graph and the existence of a 2-factor which includes or excludes a given pair of edges. Chapter 2 examines the relation between the edge-connectivity of a regular graph and the existence of a 2-factor which again includes and excludes respectively two given sets of edges. Finally in chapter 3, we obtain a sufficient condition for the square of a graph G to have a k-factor. This sufficient condition is expressed in terms of the minimum degree of G and generalizes an already known result.Τεκμήριο Εφαρμοσμένη επιχειρησιακή έρευνα στην πληροφορική: παιγνιοθεωρητικά υποδείγματα για δίκτυα χρηστών του διαδικτύουΒασιλάκης, Δημήτριος-Κωνσταντίνος; Οικονομικό Πανεπιστήμιο Αθηνών, Τμήμα Πληροφορικής; Μαγείρου, ΕυάγγελοςΗ διατριβή παρουσιάζει και αναλύει πρωτότυπα ποσοτικά υποδείγματα με στόχο τη μελέτη της διαδραστικής σχέσης μεταξύ ομάδων ή μεμονωμένων μελών σε δίκτυα χρηστών του διαδικτύου. Ως εργαλείο επίλυσης των υποδειγμάτων και πρόβλεψης της συμπεριφοράς των χρηστών χρησιμοποιείται η Θεωρία Παιγνίων. Στο πρώτο μέρος της διατριβής παρουσιάζεται ένα παιγνιοθεωρητικό υπόδειγμα που περιγράφει τη διαδραστική σχέση μεταξύ των χρηστών ηλεκτρονικής αλληλογραφίας και των αποστολέων ανεπιθύμητων μηνυμάτων (spam e-mail), όταν παρεμβάλλονται φίλτρα ανεπιθύμητης αλληλογραφίας. Το παίγνιο που προκύπτει επιλύεται με μία πρωτότυπη διαγραμματική μέθοδο για τον προσδιορισμό σημείων ισορροπίας κατά Nash 2xM παίγνια μη μηδενικού αθροίσματος. Η προτεινόμενη μέθοδος εφαρμόζεται για να αποδειχθούν θεμελιώδη θεωρήματα της Θεωρίας Παιγνίων 2xM παίγνια με διαγραμματικό τρόπο. Με βάση τα σημεία ισορροπίας του παιγνίου ηλεκτρονικής αλληλογραφίας, εκτιμάται το ποσοστό των ανεπιθύμητων μηνυμάτων σε ισορροπία και προτείνεται η βέλτιστη, από πλευράς χρηστών, ρύθμιση των φίλτρων. Στη συνέχεια το βασικό υπόδειγμα επεκτείνεται με την ανάλυση ενός μηχανισμού αποφυγής της ανεπιθύμητης αλληλογραφίας μέσω αποδείξεων ανθρώπινης αλληλεπίδρασης (human interactive proofs). Τέλος, παρουσιάζεται και επιλύεται ένα παίγνιο σημάνσεων μεταξύ των αποστολέων και των παραληπτών της ηλεκτρονικής αλληλογραφίας, εφαρμόζοντας ένα μηχανισμό ομολόγων προσοχής (attention bonds). Το δεύτερο μέρος της διατριβής αναπτύσσει ένα ολοκληρωμένο υπόδειγμα για τη συμπεριφορά ορθολογικών οντοτήτων σε ένα δίκτυο ομότιμων οντοτήτων (peer-to-peer network), στο οποίο διαμοιράζεται κάποιος πόρος. Κεντρική έννοια της ανάλυσης είναι ο αλτρουισμός των οντοτήτων, ένα ενδογενές κίνητρο που διαμορφώνει τη συνεισφορά αυτών. Προσδιορίζονται σημεία ισορροπίας, τα οποία παρέχουν μία πρόβλεψη για το επίπεδο της συνεισφοράς των οντοτήτων στο δίκτυο ως συνάρτηση της κατανομής του αλτρουισμού. Εξετάζοντας δύο διαφορετικές θεωρήσεις για τον αλτρουισμό, αποδεικνύεται ότι οι οντότητες ενδέχεται να συνεισφέρουν στο δίκτυο ακόμη και χωρίς να απαιτούνται εξωγενή κίνητρα. Το υπόδειγμα επεκτείνεται με την υιοθέτηση ενός απλού αλλά δίκαιου μηχανισμού κινήτρων, ως αποτέλεσμα του οποίου η συνολική συνεισφορά στο δίκτυο αυξάνεται.Τεκμήριο Energy efficient and adaptive service advertisement, discovery and provision for Mobile Ad Hoc NetworksVerveridis, Christopher N.; Athens University of Economics and Business, Department of Informatics; Polyzos, GeorgeMobile Ad Hoc Networks (MANETs) are networks of mobile computing nodes (e.g. portable computers, PDAs etc.) equipped with wireless interfaces and communicating with each other without relying on any infrastructure. In these networks each mobile node may act as a client, a server and a router. MANETs have emerged to fulfill the need for communication of mobile users in locations where deploying a network infrastructure is impossible, or too expensive, or simply is not available at that time. Characteristic scenarios for MANETs are disaster relief operations, battlefields and locations where infrastructure-based WLAN coverage (also called hotspots) is not provided and wireless WANs (e.g. GPRS/UMTS) are too expensive to use or too slow. Most of the research on MANETs has focused on issues dealing with the connectivity between mobile nodes in order to cope with the dynamism of such networks and the arising problems thereof. This dynamism is due to the mobility of nodes, the wireless channel’s adverse and fast changing conditions and the energy limitations of mobile nodes, all of which lead to frequent disconnections and/or node failures. These research efforts have led to the creation of a sound technical basis for dealing with the aforementioned problems regarding node connectivity in MANETs (mainly through routing protocols, link layer protocols etc.). However, solving the problems of connectivity alone is not sufficient for the adoption of MANETs. Since their basic role is to allow mobile users to exchange data and use each other’s services, there is also a need for architectures, mechanisms and protocols for Service Discovery and Provision. Service Discovery is defined in general as the process allowing networked entities to: i) advertise their services, ii) query about services provided by other entities, iii) select the most appropriate services and iv) invoke the services. Service Discovery has been mainly addressed in the context of wired networks in the past. In that context, clients and especially servers are typically powerful and resource-rich machines connected to the wired network. Service discovery and advertisement for wired networks generally follows a centralized or semi-centralized architecture assuming that well-known and “always on” service registries (or directories) exist for matching service requests to available services. Most of these architectures also rely on inefficient flooding techniques for service discovery and advertisement since resource scarcity is not a key issue. All the characteristics mentioned above, render those architectures inappropriate for service discovery in MANETs. In this context, the requirements are radically different than those of wired networks. In MANETs both clients and servers are more lightweight devices with limited resources. The assumption of (possibly dedicated and) always-available nodes serving as service directories is no longer valid. Also, frequent disconnections are a major issue affecting service availability. Disconnections may happen due to node mobility, due to depletion of the energy resources of the nodes, or even due to switching-off of some nodes. It is rather straightforward that in such environments it is of utmost importance to have energy efficient service discovery mechanisms. In this thesis we develop two energy efficient service discovery protocols integrated into the routing process in order to avoid redundant network messaging. Both developed protocols are distributed in nature and employ both proactive and reactive information dissemination techniques. The performance of the developed protocols is thoroughly investigated through extensive simulations in terms of energy consumption, service discoverability and achievable service availability under a wide range of settings (e.g. mobility, network density, channel characteristics, service replication, traffic patterns). The first (basic) protocol, called Extended Zone Routing Protocol (E-ZRP) is used in our study for comparing cross-layer with application-layer based service discoveryprotocols. Although integrating service and route discovery has been already proposed in the past, previous work regarding the energy gains of implementing service discovery at the routing layer as compared to implementing it at the application layer (as a separate process) is problematic due to the following reasons: i) most approaches either deal with energy only implicitly (by measuring the overhead in number of packets) or employ non-realistic energy consumption modeling, ii) the majority of proposed approaches involve performance analysis comparisons of the developed route and service discovery protocols against unrealistic application layer based service discovery protocols which are based on global flooding and hence are not suitable to MANETs, and iii) none of those approaches takes into consideration the impact of protocol message sizes on protocol performance. In our work we have performed fullstack simulations employing a realistic energy consumption model, accounting for the actual energy consumption of the nodes, in order to reveal the real gains from integrating the service discovery process with the routing process. The integrated service discovery protocol, named E-ZRP proves to be much more efficient compared to a similar but application layer based service discovery protocol. More importantly, we developed an even more sophisticated service discovery protocol, called Adaptive SerVicE and Route Discovery ProTocol for MANETs (AVERT). AVERT demonstrates superior efficiency than the best alternatives, mainly due to its built-in capability of adapting to a volatile environment such as a MANET. Using a monitoring mechanism for traffic seen locally, a node may adapt the operation of AVERT to best match the fluctuating traffic and mobility conditions typically found in MANETs. In parallel to the aforementioned research effort, we developed a novel mechanism for service providers to select to serve those clients from a MANET that can maximize their profits. Optimized service provisioning is a challenging problem in dynamic environments such as MANETs. We consider the nodes in MANETs to be independent, rational agents trying to maximize their profits through service provision. We model this problem as a Generalized Assignment Problem (GAP). We adopt a pay-as-you-go model, where clients pay for the service as long as they are receiving the service, since a pay-in-advance model would be unfair especially in MANETs where connection loss is very probable. We introduce into the proposed profit maximization algorithm expected payoffs based on estimates of server-to-client connectivity. Those estimations can be used for computing the actual payoff that will be received from any client that is selected by the service provider. We experimentally study cases with non-cooperative and cooperative servers and investigate the gain of the estimate based profit maximization algorithm versus a classic profit maximization algorithm, which does not take into account the network’s dynamics that affect server-to-client connectivity. We especially study the duration of connectivity (irrespectively of path changes) between two nodes of a MANET. Previous work, both analytical and experimental, has focused only on estimating the duration of a single path between a client and a server without considering changes to a path’s length caused by node mobility. The connectivity between two nodes irrespectively of changes in paths or the path length has not been investigated. In this dissertation we derive an approximation of the connectivity duration, taking into account the network’s density (number of nodes, terrain size, wireless transmission range), the node speed and the initial distance (in number of hops) between two nodes (one representing a client and the other a server). Simulations show that our approach achieves up to three-fold improved server profits compared to the classical one and is especially suited for MANETs with high-mobility and/or low density, which verifies that the proposed model accurately captures the effects of server to client connectivity on the overall performance. Summing up, in this thesis we propose energy efficient highly adaptive service discovery protocols integrated with hybrid routing protocols. We implemented and experimentally evaluated those protocols to show their energy and service discoverability/availability gains under different MANET conditions and against similar application layer-based as well as routing layer-based protocols. We also developed and evaluated an innovative mechanism for allowing mobile service providers to maximize their profits when offering their service in MANETs, by estimating their connectivity to candidate clients.Τεκμήριο Linear and non linear dimensionality reduction for distributed knowledge discoveryMagdalinos, Panagis; Athens University of Economics and Business, Department of Informatics; Vazirgiannis, MichalisAn increasing number of contemporary applications produce massive volumes of very high dimensional data. In scientific databases, for example, it is common to encounter large sets of observations, represented by hundreds or even thousands of coordinates. Unfortunately the rate of data generation and accumulation significantly outperforms our ability to explore and analyze it. Nevertheless, in order to extract knowledge from these datasets, we need to access the underlying, hidden information. However, the size and complexity of these collections makes their processing and analysis impractical or even ineffective [13, 47]. Therefore, scaling up knowledge discovery algorithms for data of both high dimensionality and cardinality has been recently recognized as one of the top-10 problems in data mining research [95].In parallel, the evolution of the internet as well as the emergence of novel applications, such as peer-to-peer systems, has led to an unprecedented distribution of available information. Data is dispersed among network nodes, making the cost of centralizing and subsequent processing prohibitive. Consequently, distributed data mining and distributed knowledge discovery have also emerged as highly challenging tasks[95]. Nevertheless, the vast amount of generated data dictates methods that are fast, exhibit low requirements in terms of computational resources and can be applied to various network setups. Motivated by the previous analysis, this thesis attempts to provide a solution through the definition of efficient and effective dimensionality reduction algorithms. The proposed methods exhibit minor requirements in terms of computational resources without compromising the quality of the produced results; therefore can be exploited in the context of centralized and distributed preprocessing for knowledge discovery. Towards this end,• We introduce FEDRA (Chapter 3, [62, 63]), a dimensionality reduction algorithm which poses minimum time and space requirements and is ideal for large FEDRA is an acronym for A Fast and Efficient Dimensionality Reduction Algorithm datasets of particularly high cardinality and dimensionality.• Inspired by the nature of landmark based dimensionality reduction algorithms(Chapter 2) we introduce the distributed adaptation of FEDRA ([62, 61]) and extend its underlying methodology in order to derive a framework for the decentralization of any landmark based dimensionality reduction algorithm (Chapter 3, Section 3.4)• We propose a distributed non linear dimensionality reduction algorithm, the Distributed Isomap (Chapter 4, [66, 65]) which to the best of our knowledge comprises the first of this kind. Additionally, motivated by recent research results on text-mining ([41, 17, 101, 78, 71]) we propose its application for hard dimensionality reduction problems related with text-mining.• Finally, we introduce X-SDR (Chapter 5, [64]), a prototype that enables the integration and evaluation of any dimensionality reduction algorithm. X-SDRis an open source tool that supports the evaluation of methods through experimentation on artificial and real world datasets thus promoting itself as an ideal candidate platform for research and teaching in academia.Τεκμήριο Stability of spectral learning algorithms: theory, methodologies and applicationsMavroeidis, Dimitrios; Athens School of Economics and Business, Department of Informatics; Vazirgiannis, MichalisIn Data Mining an important research problem is the identification and analysis of theoretical properties that characterize and explain the behavior of learning algorithms. Based on such theoretical tools the comparison and analysis of algorithms can be based on rigorous and sound criteria other than simply their empirical behavior. In this context, an important criterion that is widely used for assessing the quality of learning algorithms is stability. Stability essentially evaluates the sensitivity of the output of a learning algorithm with respect to small perturbations of the input. An algorithm is said to be stable if it is insensitive to perturbations and unstable if even small perturbations of the input can significantly alter the algorithm’s output. Based on this definition it is natural to require that an algorithm is stable. A learning paradigm that has been recently developed in the context of Data Mining considers the use of Spectral techniques for addressing several data mining tasks, ranging from Dimensionality Reduction to Clustering and Classification. Spectral algorithms are commonly perceived as approximate solutions to certain combinatorial optimization problems and derive their solutions based on the eigenvectors of appropriate input matrices. The stability of the results depends on certain algebraic properties of these matrices, thus the study of stability reduces to the study of these algebraic properties.In this thesis we focus on the stability of spectral learning algorithms and consider the following research questions: How can we measure the stability of spectral learning algorithms? What is the relation of sample size and (in)stability? What is the relation between stability and efficiency? How can we identify the sufficient sample size for guaranteeing stability? How can we enhance the stability of spectral learning algorithms? How can the concept of stability be employed in distributed learning algorithms? Based on the aforementioned research directions, the contributions of this thesis can be summarized in the following. We define an efficient bootstrap based methodology for measuring the stability of Principal Components Analysis (PCA) [88], Spectral k-Means and Normalized Spectral Clustering [85]. • We propose a Feature Selection Framework that enhances the stability of Principal Components Analysis PCA [88] and Spectral Clustering/Ordering [83]. • We propose a Semi-Supervised Framework that enhances the stability and efficiency of Spectral Ordering[83, 84]. • Motivated by [87] where highly accurate classification accuracy is achieved with very small training data sets, we associate the concept of sampling variability to instability and Sample Size, and propose a Spectral Clustering algorithm that automatically identifies the sufficient sample size that guarantees stability [85]. • We define a Distributed Spectral Clustering framework, that aims in minimizing instability. The proposed framework is demonstrated to obtain highly accurate models with low bandwidth consumption [85].Τεκμήριο Equilibrium and Optimal Control Policies for Queueing Models with Variable Service CapacityDimitrakopoulos, Yiannis; Athnes University of Economics and Business, Department of Informatics; Mageirou, Evangelos; Burnetas, ApostolosIn this thesis, we develop and analyze stochastic queueing models with which we study the effect of a varying service rate policy on customer arrival rate, considering two different ways for controlling customer entrance. In the first problem, we assume that customers respond strategically to excessive delays by taking into account the trade-off between service benefit and expected delay in the presence of a service control dynamic policy employed by the system administrator. In this case, arrival rate control is implicit, since it is imposed in equilibrium in the game between potential incoming customers. The effect of the service rate policy is that the possible equilibria are generally more than one. In the second problem, we assume that the system administrator controls both customer arrival rate and the service rate employed with respect to the system congestion in a stochastic dynamic framework. More specifically, in the first part of this dissertation, we consider the problem of customer equilibrium behavior on joining a single server Markovian queueing system with varying service rate between a low and a high value, according to a threshold based dynamic policy. We assume that arriving customers decide whether they will join a queue and receive a fixed reward or balk, i.e., decline to enter the queue, and compromise with an alternative payment. When customers make this decision they don't have any information on the system congestion upon arrival, but they know the implemented service policy. Since a tagged customer's expected delay is not monotone increasing in the arrival rate, the equilibrium strategy is not unique, and we show that the number of such strategies can vary from zero to three. Moreover, regarding the problem of maximizing the overall customer expected net benefit, we show numerically that, the optimal arrival rate that maximizes the customer's overall welfare lies between the extreme equilibrium arrival rates. Since the equilibrium arrival rates and the social optimal strategies do not coincide, we analyze this inefficiency considering the Price of Anarchy. We show that as the value of balking decreases, the price of anarchy becomes unbounded since customers are indifferent between joining and balking under any equilibrium strategy. The latter problem of social optimization can be viewed as one way of centrally controlling both admission and service rate. A more realistic setting to examine the impact of the service rate switch to a higher value on arrival rate is to assume that the system manager observes the queue length and dynamically sets both the arrival and the service rate. In the second part of this dissertation, we consider the effect of varying the service rate between a finite set of increasing values in an M/M/1 queue with admission control. Here, we formulate a Markov Decision Process model to analyze the joint admission and service control, considering both discounted and average expected profit maximization with the service payment being made upon admission or at the end of service. For each case, we derive a thresholdbased optimal policy for both controls. We define the value of service rate exibility as the difference in optimal profit between the joint control model and an admission control model with service rate set at the lowest value. Under a simple condition between the service payment and the relative cost of service rate increase, we show that the value of service rate exibility is positive, and, increasing in the system congestion. Moreover, the optimal admission threshold increases, so, more customers are accepted due to service rate exibility. Finally, sensitivity analysis with respect to the service payment for the different cases is pursued numerically.Τεκμήριο Organizing and Searching Data in Unstructured P2P NetworksDoulkeridis, Christos; Athens University of Economics and Business, Department of Informatics; Vazirgiannis, MichalisAs data generation becomes increasingly inherently distributed, either due to usergenerated (multimedia) content or because of application-specific needs (sensor networks, data streams, etc.), traditional centralized architectures fail to address the new challenges of contemporary data management. A promising solution for the design and deployment of global-scale applications is the exploitation of the peer-to-peer (P2P) paradigm. P2P has emerged as a powerful model for organizing and searching large data repositories distributed over autonomous independent sources. The main topic and contribution of this thesis is the unsupervised organization of content into Semantic Overlay Networks (SONs), in a decentralized and distributed manner, and subsequently a variety of techniques for efficient searching and query processing in unstructured P2P systems. SONs have been proposed in the relevant research literature, as a way to organize peers into thematic groups, thereby enabling query routing to specific peer groups in a deliberate way, instead of blind forwarding. In particular, this work focuses on unstructured P2P networks that preserve peer autonomy. A novel protocol for unsupervised, distributed and decentralized SON construction is proposed, named DESENT [35, 38], which employs distributed clustering of peer contents, respecting the requirements imposed by the distributed nature of the environment [138]. Exploiting the generated SONs, we propose efficient routing strategies for answering similarity search queries [37, 39]. The approach is applied and tested in a distributed IR setting, aiming to address some of the limitations of P2P IR/web search. Towards this goal, a distributed dimensionality reduction algorithm is proposed [96], in order to reduce the high-dimensional feature space and improve clustering quality. Assuming a super-peer architecture we propose an approach called SIMPEER [43] that efficiently supports similarity search over data distributed over a large set of peers. We show how range queries and nearest neighbor queries can be processed. We also explore how to support non-traditional queries (such as top-k [141] and skylines [139]) that involve ranking. Furthermore, by relaxing the restriction of completely unsupervised environment and assuming a semi-supervised context, a novel technique for P2P summary caching of hierarchical information is presented, exploiting either predefined taxonomies [104] or XML schema information [36, 40], which is applied in mobile P2P context-aware environments to improve query routing [45, 44].Τεκμήριο Απόδοση δικτυακών πόρων μέσω οικονομικών μηχανισμών δημοπρασιώνΔραμιτινός, Εμμανουήλ; Οικονομικό Πανεπιστήμιο Αθηνών, Τμήμα Πληροφορικής; Σταμούλης, ΓεώργιοςΗ ραγδαία ανάπτυξη του Διαδικτύου (Internet) τα τελευταία χρόνια συνεπάγεται τη συνεχώς αυξανόμενη ανάγκη για εύρος ζώνης τόσο στα σημεία πρόσβασης του τελικού χρήστη στο δίκτυο, όσο και στο δικτυακό κορμό. Λόγω ανταγωνισμού, οι πάροχοι των σταθερών δικτύων είναι αναγκασμένοι πλέον να προσφέρουν συμβόλαια για μικρές χρονικές κλίμακες σε ανταγωνιστικές τιμές, οι οποίες ταυτόχρονα να αντικατοπτρίζουν το φόρτο του δικτύου τους και τη ζήτηση της αγοράς για τις υπηρεσίες τους. Έτσι, υπάρχει ολοένα αυξανόμενο ενδιαφέρον για νέα δυναμικά συστήματα αγοραπωλησίας εύρους ζώνης καθώς και αποτελεσματικότερων τρόπων διαχείρισης και χρέωσης των δικτυακών πόρων. Παράλληλα, η αλματώδης ανάπτυξη των ασύρματων δικτύων, συμπεριλαμβανομένων των κυψελοειδών, και η αύξηση της ζήτησης των αντιστοίχων υπηρεσιών, έχει οδηγήσει σε ιδιαιτέρως έντονη ζήτηση για δέσμευση φάσματος σε συγκεκριμένες ζώνες συχνοτήτων, προκειμένου να καταστεί εφικτή η παροχή των υπηρεσιών 2ης (GSM), 2,5 (GPRS) και 3ης γενεάς (UMTS). Το διαθέσιμο φάσμα του ασυρμάτου δικτύου πρόσβασης (Radio Access Network) των δικτύων αυτών είναι εξαιρετικά περιορισμένο. Αντίθετα, οι νέες υπηρεσίες που θα εξυπηρετούνται από τα δίκτυα αυτά, έχουν σημαντικά υψηλότερες απαιτήσεις από τα δίκτυα 2ης γενεάς: Τα μηνύματα με πολυμεσικό περιεχόμενο (MMS) αντικαθιστούν τα σύντομα γραπτά μηνύματα (SMS), ενώ η μετάδοση φωνής συνυπάρχει με τη μετάδοση φωτογραφιών, αρχείων και video σε πολλαπλάσιους ρυθμούς μετάδοσης. Οι υπηρεσίες αυτές έχουν χρονική διάρκεια κατά πολλές τάξεις μεγέθους μεγαλύτερη από αυτή στην οποία δεσμεύει πόρους το δίκτυο (π.χ. τα UMTS frames έχουν διάρκεια 10 msec). Παράλληλα, η μελλοντική σύγκλιση των ανωτέρω τεχνολογιών με άλλες τεχνολογίες πρόσβασης και η δυνατότητα εξυπηρέτησης πολυμετάδοσης (multicast), κάνουν το τηλεπικοινωνιακό περιβάλλον ακόμα πιο δυναμικό. Από τα παραπάνω, γίνεται φανερό ότι είναι απαραίτητη η εφαρμογή μηχανισμών δέσμευσης πόρων και χρέωσης αυτών οι οποίοι να είναι κατάλληλοι για αυτό το δυναμικό τηλεπικοινωνιακό περιβάλλον. Ένας πολύ δημοφιλής τρόπος ανταλλαγής αγαθών για δυναμικές οικονομίες, ο οποίος βασίζεται στην έννοια του ανταγωνισμού, είναι οι δημοπρασίες (auctions). Οι δημοπρασίες εν γένει παρέχουν έναν απλό, ταχύ και διαφανή μηχανισμό καθορισμού των τιμών των διαφόρων αγαθών. Το πλέον χαρακτηριστικό και επιτυχημένο δείγμα εφαρμογής τέτοιου μηχανισμού αποτελεί η δημοπρασία τύπου FCC που χρησιμοποιήθηκε για την εκχώρηση ζωνών φάσματος σε εταιρείες κινητής τηλεφωνίας. Αντικείμενο της παρούσας διδακτορικής διατριβής αποτελεί η χρήση δημοπρασιών για την εκχώρηση δικτυακών πόρων.Αρχικά, πραγματοποιείται λεπτομερής μελέτη της δημοπρασίας τύπου FCC. Συγκεκριμένα, παρουσιάζονται οι βασικές στρατηγικές παικτών που εφαρμόσθηκαν στην πράξη σε δημοπρασίες τέτοιου τύπου. Στη βιβλιογραφία, για την περιγραφή των στρατηγικών των παικτών, χρησιμοποιούνται ασαφείς όροι οι οποίοι καθιστούν απαραίτητη την ύπαρξη μιας κοινής και αυστηρής γλώσσας για τον καθορισμό τέτοιων στρατηγικών. Γι’ αυτό το λόγο στην παρούσα διατριβή προτείνεται μια τέτοια «γλώσσα», μία κανονική LR(1) γραμματική. Η επιλογή αυτή εξασφαλίζει την αυστηρότητα και πληρότητα της περιγραφής και την αναγνωσιμότητα της στρατηγικής ακόμα και από απλά προγράμματα (αυτόματα). Παράλληλα, η εν λόγω δημοπρασία χαρακτηρίζεται από έλλειψη προδιαγεγραμμένου σημείου ισορροπίας Nash στο οποίο πρόκειται να καταλήξει. Επομένως, η προδιαγραφή μιας επιτυχημένης στρατηγικής είναι ιδιαίτερα σημαντική, γι’ αυτό προτείνεται και αποτιμάται μια τέτοια στρατηγική. Τέλος, έχει υλοποιηθεί και μια πλατφόρμα λογισμικού η οποία επιτρέπει τη διεξαγωγή τέτοιου τύπου δημοπρασιών για εκπαιδευτικούς λόγους ή για τη διερεύνηση της αποτελεσματικότητας των στρατηγικών κάτω από μια πλειάδα συνθηκών αγοράς. Επίσης, παρέχει λειτουργίες συστήματος στήριξης αποφάσεων σε παίκτες που συμμετέχουν στη δημοπρασία ως προς τις προσφορές οι οποίες πρέπει να υποβληθούν βάσει μιας προκαθορισμένης στρατηγικής. Το επίκαιρο πρόβλημα της δέσμευσης πόρων σε σταθερά δίκτυα γενικής τοπολογίας μέσω δημοπρασιών επίσης αποτελεί αντικείμενο μελέτης της παρούσας διατριβής. Συγκεκριμένα, παρουσιάζεται και αναλύεται ο μηχανισμός MIDAS ο οποίος είχε σε αρχική μορφή εισαχθεί σε προηγούμενη εργασία. Πρόκειται για ένα μηχανισμό δημοπράτησης του εύρους ζώνης δικτύων και βασίζεται στη διεξαγωγή ταυτόχρονων δημοπρασιών μειούμενης τιμής (δηλαδή Ολλανδικών δημοπρασιών) πολλαπλών μονάδων σε κάθε σύνδεσμο του δικτύου. Το εύρος ζώνης όλων των συνδέσμων ενός δικτύου δημοπρατείται περιοδικά και εκχωρείται στους νικητές της δημοπρασίας για μία συγκεκριμένη χρονική περίοδο. Πρωταρχικός σκοπός του μηχανισμού MIDAS είναι η εκχώρηση του εύρους ζώνης στους χρήστες του δικτύου που το αποτιμούν περισσότερο, δηλαδή η μεγιστοποίηση της κοινωνικής ευημερίας. Ο μηχανισμός παρέχει τα κατάλληλα κίνητρα στους παίκτες ώστε οι τελευταίοι να αποστέλλουν ειλικρινείς προσφορές (truthful bids). Δεδομένης αυτής της ιδιότητας συμβατότητας κινήτρων, στην παρούσα διατριβή αποδεικνύεται ότι ο μηχανισμός MIDAS αποδίδει το εύρος ζώνης βέλτιστα σε όλο το δίκτυο αν ο δημοπράτης διαθέτει πλήρη πληροφόρηση (full information) για τη ζήτηση. Εάν διαθέτει μερική πληροφόρηση, αποδεικνύεται ότι και πάλι η εκχώρηση του εύρους ζώνης γίνεται με τέτοιο τρόπο ώστε η τιμή της επαγόμενης κοινωνικής ευημερίας να είναι κοντά στη μέγιστη δυνατή. Επίσης, ο μηχανισμός μπορεί να επιτύχει μεγιστοποίηση των εσόδων του πωλητή, αν δεν πωληθεί ολόκληρη η διαθέσιμη ποσότητα εύρους ζώνης του δικτύου. Η πολυπλοκότητα του προβλήματος της ακριβώς βέλτιστης (ως προς τη μεγιστοποίηση της κοινωνικής ευημερίας) εκχώρησης του εύρους ζώνης είναι NP-hard. Το μέγιστο απαιτούμενο πλήθος βημάτων εκτέλεσης του εν λόγω μηχανισμού είναι μικρό και ανεξάρτητο του πλήθους των χρηστών και συνδέσμων του δικτύου. Με αυτόν τον τρόπο, ο μηχανισμός MIDAS αποτελεί μια γρήγορη σχεδόν βέλτιστη λύση σε ένα εν γένει πολύπλοκο (NP-hard) πρόβλημα βελτιστοποίησης.Η μεγάλη ανάπτυξη των ασύρματων κυψελοειδών δικτύων 3ης γενεάς σε συνδυασμό με την αυξανόμενη ανάγκη για ορθή και αποτελεσματική διαχείριση του περιορισμένου φάσματος του ασύρματου δικτύου πρόσβασης καθιστά ιδιαίτερα επίκαιρη τη μελέτη της εκχώρησης των πόρων μέσω δημοπρασιών. Στην παρούσα διατριβή προτείνεται ο μηχανισμός ATHENA: Πρόκειται για ένα μηχανισμό εκχώρησης πόρων για ένα δίκτυο 3ης γενεάς-UMTS με τέτοιο τρόπο ώστε οι δικτυακοί πόροι να μοιράζονται κατά οικονομικά βέλτιστο τρόπο ανάμεσα στους πελάτες και η χρέωση να αντικατοπτρίζει το φόρτο του δικτύου. Ο μηχανισμός ο οποίος προτείνεται αποτελείται από α) μια σειρά επαναλαμβανόμενων δημοπρασιών πολλαπλών μονάδων τύπου Generalized Vickrey Auction, μία για κάθε χρονοθυρίδα, και β) προκαθορισμένες συναρτήσεις αυτόματης υποβολής προσφορών στις ανωτέρω δημοπρασίες, οι οποίες επιλέγονται και παραμετροποιούνται από το χρήστη. Οι συναρτήσεις αυτές εκφράζουν τις προτιμήσεις χρηστών με απαίτηση για σταθερό ρυθμό μετάδοσης δεδομένων τύπου CBR όσον αφορά στις μορφές εκχώρησης των πόρων όταν η συνεπής εκχώρηση του ρυθμού αυτού δεν είναι εφικτή και λαμβάνουν υπ’ όψιν το ιστορικό της εκχώρησης πόρων στην παρούσα σύνοδο του χρήστη. Αυτό το γεγονός αποτελεί μια πρωτότυπη συμβολή της παρούσας διατριβής, η οποία μάλιστα έχει υιοθετηθεί εσχάτως και από άλλους ερευνητές. Η πειραματική αποτίμηση του μηχανισμού καταδεικνύει ότι ο μηχανισμός ATHENA έχει τη σημαντική ιδιότητα ότι η πλειονότητα των μορφών εκχώρησης των πόρων των χρηστών είτε είναι σχεδόν συνεπείς είτε περιλαμβάνουν εξαιρετικά περιορισμένες ποσότητες μονάδων πόρων για τις οποίες ο χρήστης χρεώνεται ελάχιστα. Ο μηχανισμός παρέχει τα ορθά κίνητρα στους παίκτες για τη χρήση του δικτύου. Ο μηχανισμός αποτιμάται τόσο θεωρητικά μέσω παιγνιοθεωρητικής ανάλυσης, όσο και πειραματικά μέσω λογισμικού που υλοποιήθηκε γι’ αυτόν το σκοπό. Επίσης, πραγματοποιείται η συγκριτική αποτίμηση του μηχανισμού με δύο διαδεδομένες μη οικονομικές πολιτικές εκχώρησης πόρων, τον αλγόριθμο κυκλικής εκχώρησης πόρων (Round Robin) και τον αλγόριθμο εξυπηρέτησης κατά σειρά άφιξης (First Come First Served) ως προς την ικανοποίηση των χρηστών από τις ληφθείσες υπηρεσίες. Η πειραματική αποτίμηση καταδεικνύει ότι ο προτεινόμενος μηχανισμός δημοπρασίας είναι σαφώς προτιμότερος τόσο από τον αλγόριθμο Round Robin ο οποίος συνεπάγεται πολλές ενδιάμεσες μορφές εκχώρησης πόρων, όσο και από τον FCFS λόγω της μεγάλης καθυστέρησης εισόδου στο δίκτυο και της χαμηλότερης επαγόμενης κοινωνικής ευημερίαςΠαράλληλα, πραγματοποιήθηκε ανάλυση μεταβατικών φαινομένων του μηχανισμού ATHENA και της συντελούμενης απόδοσης πόρων στους νικητές μέσω του ορισμού και της ανάλυσης ενός θεωρητικού υποδείγματος τυχαίου περιπάτου που αντιστοιχεί στην αυξομείωση της τιμής κατωφλιού της δημοπρασίας. Το υπόδειγμα χρησιμοποιείται για να αναλυθούν οι προκύπτουσες μορφές εκχώρησης πόρων για παίκτες των οποίων η οριακή χρησιμότητα και συνακόλουθα η αποστελλόμενη προσφορά στη δημοπρασία παραμένει σταθερή όσο ο παίκτης κερδίζει, ενώ μειώνεται εκθετικά με το χρόνο όταν η τιμή κατωφλιού της δημοπρασίας είναι μεγαλύτερη από αυτήν. Συγκεκριμένα, για έναν εν γένει ανταγωνιστικό παίκτη μελετώνται: α) Η πιθανότητα, δεδομένου του ότι έχασε σε κάποια δημοπρασία, να ξανακερδίσει κάποια στιγμή στο μέλλον, και β) το αναμενόμενο πλήθος των δημοπρασιών στις οποίες οι προσφορές του παίκτη δεν θα υπερβαίνουν την τιμή της δημοπρασίας. Χρησιμοποιώντας το υπόδειγμα αυτό, κάθε παίκτης της δημοπρασίας δύναται να εξετάσει την επίδραση της τιμής του παράγοντα εκθετικής μείωσης της συνάρτησης υποβολής προσφορών του στην αναμενόμενη ποιότητα υπηρεσίας. Τέλος καθορίζεται πώς μπορεί να επεκταθεί ο μηχανισμός ΑΤΗΕΝΑ ώστε να υποστηρίζει τη δημοπράτηση πόρων και για συνδέσεις πολυμετάδοσης. Συγκεκριμένα, εξετάζονται οι ιδιότητες του εν λόγω μηχανισμού όταν οι ανταγωνιζόμενες ροές δεν είναι μόνο ατομικές ροές χρηστών (unicast flows), αλλά παρέχεται και η δυνατότητα εξυπηρέτησης ροών πολυμετάδοσης (multicast flows), των οποίων εξετάζεται η επίδραση στην κοινωνική ευημερία και στα έσοδα του πωλητή. Η ανάλυση αυτή εστιάζει στην περίπτωση κατά την οποία οι συμμετέχοντες έχουν σημαντικές διαφορές ως προς το χρηματικό ποσό το οποίο είναι διατεθειμένοι να καταβάλλουν για την υπηρεσία (wealth asymmetry). Επίσης εξετάζεται η επίπτωση της πολυμετάδοσης στις μορφές εκχώρησης πόρων των χρηστών. Προκύπτει ότι η πολυμετάδοση βελτιώνει την κοινωνική ευημερία και σε πολλές περιπτώσεις και τα έσοδα. Συνοψίζοντας, η παρούσα διδακτορική διατριβή αποτελεί μία μελέτη δυνατότητας εφαρμογής οικονομικών μηχανισμών και ιδιαίτερα δημοπρασιών για την εκχώρηση των πόρων σταθερών και κυψελοειδών δικτύων στους ανταγωνιζόμενους χρήστες τους. Προτείνονται και μελετώνται οικονομικοί μηχανισμοί για διάφορα προβλήματα με θεωρητικό ενδιαφέρον αλλά και πρακτική σημασία. Μέσω θεωρητικής και πειραματικής αποτίμησης αποδεικνύεται ότι οι μηχανισμοί αυτοί αποτελούν ταχείς, αποδοτικές, και πρακτικά εφαρμόσιμες λύσεις.Τεκμήριο The contribution of experiential and material purchases to happiness: expanding the experience recommendation(2018) Kousi, Sofia; Baltas, George; Indounas, Konstantinos; Andronikidis, Andreas; Dimitriadis, Sergios; Siοmkos, George J.; Skarmeas, Dionysios; Kokkinaki, FloraThe experience recommendation posits that experiential purchases, such as concert and airplane tickets, contribute more to happiness than material purchases, such as clothes and cars. Closer examination of the extant research on the topic and the empirical explanations provided for this effect, reveals underlying asymmetries in the nature of these purchases, which contribute to the happiness advantage of experiential purchases. As such, the experience recommendation seems to only partially explain the relationship between consumption and happiness, thus limiting the value of any recommendation for maximizing happiness from spending. Τhe present research proposes and examines two asymmetries which give experiential purchases a happiness advantage. Specifically, the present study hypothesizes that experiential purchases are inherently more hedonic and self-expressive than material purchases. More importantly, it hypothesizes that once these asymmetries are corrected, material purchases elicit equal levels of happiness as experiential purchases. Through a qualitative study and five experiments, it is demonstrated that recalled experiential purchases tend to be more hedonic and self-expressive than material purchases. These characteristics give experiential purchases a happiness advantage, as they mediate the relationship between happiness and purchase type. The results of the present research contribute to the refinement of the experience recommendation, and indicate that under specific circumstances, material purchases can also make people happy. These findings refine the experience recommendation, deepen our understanding of the relationship between consumption and happiness, and add to a growing volume of research that reveals the complexity of this relationship. The data points to a necessary shift away from the question of whether experiential or material purchases make people happier, to asking what elements and characteristics of any purchase are likely to elicit happiness. The results suggest a prioritization of highly self-expressive and hedonic purchases, as these two characteristics positively influence the elicited happiness.Τεκμήριο Πλαίσιο διαχείρισης κατανεμημένων συστημάτων και υπηρεσιών βασισμένο σε τεχνολογίες ιστού: εφαρμογή στην περίπτωση των δικτύων διανομής περιεχομένουΟικονόμου, Γεώργιος; Οικονομικό Πανεπιστήμιο Αθηνών, Τμήμα Πληροφορικής; Αποστολόπουλος, ΘεόδωροςΗ παρούσα διατριβή πραγματεύεται ζητήματα σχετικά με τη διαχείριση κατανεμημένωνσυστημάτων και υπηρεσιών και εκπονήθηκε στο Τμήμα Πληροφορικής του ΟικονομικούΠανεπιστημίου Αθηνών. Στόχος της είναι η ανάλυση και λεπτομερής σχεδίαση μιας αρχιτεκτονικήςδιαχείρισης κατανεμημένων συστημάτων. Η ανωτέρω περιγραφή του αντικειμένου περιλαμβάνειδύο έννοιες οι οποίες δεν είναι αρκετά σαφείς και χρήζουν περισσότερης ανάλυσης. Οι έννοιεςαυτές είναι “Διαχείριση” και “Κατανεμημένο Σύστημα”.Τεκμήριο Fighting an unfair battle: unconventional defenses against Advanced Persistent ThreatsVirvilis-Kollitiris, Nikolaos; Athens University of Economics and Business, Department of Informatics; Gritzalis, DimitrisThe number and complexity of cyber-attacks has been increasing steadily in recent years. The major players in today’s cyber conflicts are well organized and heavily funded teams with specific goals and objectives, working for or supported by a nation-state. A commonly used term to describe such teams/groups is Advanced Persistent Threat (APT). APT target the communication and information systems of government, military and industrial organizations and are willing to use vast amounts of money, time and expertise to reach their goals. A clear indication of the level of sophistication of APT is their impressive arsenal. The complexity and capabilities of recently discovered malware used to facilitate such attacks are remarkable: Stuxnet, Duqu, Flame, Red October, MiniDuke and more recently Regin are examples of highly sophisticated malware, the development of which required skillful individuals – in some cases (e.g. Stuxnet) with expertise in multiple technology fields – as well as substantial financial resources. In addition, serious insider attacks have occurred that resulted in the publication of several thousand classified documents, highlighting the fact that even in sensitive institutions, the effectiveness of the existing security safeguards is insufficient. Advances in attacker sophistication have not been matched by similar defensive advances. The concept of keeping the internal, trusted network separated from the external, untrusted one (i.e. boundary protection) has become obsolete. The use of blacklists or signatures for attack detection is practically useless against sophisticated attackers. The security industry, having spent decades developing security products such as anti-malware solutions and intrusion-detection/prevention systems, refuses to admit the shortcomings of these products. It is not uncommon for security companies to advertise that their products can detect and stop APT, even though the same products have been unable to detect such attacks for several years. Furthermore, C-level executives fail to understand the need for more robust security mechanisms, as they believe that by following vendor recommendations and making significant investments in traditional security solutions, they will keep their organization secure. However reality has proven them wrong, over and over again. In order to defend against such sophisticated adversaries, it is necessary to redesign our defenses and develop technologies focused more on detection than prevention. The purpose of this thesis is to offer a comprehensive view of the APT problem by analyzing the most common techniques, tools and attack paths that attackers are using, and highlighting the shortcomings of current security solutions. The use of deception techniques for attack detection is one of the integral focal points of this thesis. Based on this concept, a novel APT detection model is proposed, implemented and evaluated. The evaluation results highlight the significant efficacy of the model in detecting sophisticated attacks, with a very low false positive rate.
- «
- 1 (current)
- 2
- 3
- »