Alexandros Karakos

Real time DDoS detection using fuzzy estimators

We propose a method for DDoS detection by constructing a fuzzy estimator on the mean packet inter arrival times. We divided the problem into two challenges, the first being the actual detection of the DDoS event taking place and the second being the identification of the offending IP addresses. We have imposed strict real time constraints for the first challenge and more relaxed constraints for the identification of addresses. Through empirical evaluation we confirmed that the detection can be completed within improved real time limits and that by using fuzzy estimators instead of crisp statistical descriptors we can avoid the shortcomings posed by assumptions on the model distribution of the traffic. In addition we managed to obtain results under a 3 sec detection window.

Homogeneity and Enrichment: Two Metrics for Web Applications Assessment

This paper presents a digital controlled distributed converter system with a switched-mode boost type PWM rectifier and several bidirectional buck/boost dc/dc converters for battery charge and discharge system using digital signal controller (DSC). The rectifier can implement bidirectional power flow, sinusoidal current control and unit power factor in grid side. The system is able to control power flow between the ac supply and dc side, or within the dc side. It is significant in development of green, economizing energy equipments. Simulation and experimental results are shown to verify the proposed distributed control scheme.Earlier studies suggest that educational institutions may further benefit from Learning Management Systems (LMSs) by generating reports regarding courses. That is, with the use of tools focus to assess students paths into course content. This paper affirms that educational meaningful information can be extracted from LMS student logged data and presents a methodology on how such findings may assist to the development of a reporting tool for educators. In the context of course assessment from student logged data, we introduce a new metric called homogeneity that extends the use of an older one, enrichment. We also propose a new algorithm that tries to qualify course content by classifying the course based on students’ course paths interest. We applied our algorithm to Open eClass LMS tracking data of an academic institution and we present the results identified in 12 courses along with interest insights. We leave for future development the creation of an automated tool of our algorithm implemented for the Open eClass platform.

A User-Centered, Object-Oriented Methodology for Developing Health Information Systems: A Clinical Information System (CIS) Example

The aim of this study is to present our perspectives on healthcare analysis and design and the lessons learned from our experience with the development of a distributed, object-oriented Clinical Information System (CIS). In order to overcome known issues regarding development, implementation and finally acceptance of a CIS by the physicians we decided to develop a novel object-oriented methodology by integrating usability principles and techniques in a simplified version of a well established software engineering process (SEP), the Unified Process (UP). A multilayer architecture has been defined and implemented with the use of a vendor application framework. Our first experiences from a pilot implementation of our CIS are positive. This approach allowed us to gain a socio-technical understanding of the domain and enabled us to identify all the important factors that define both the structure and the behavior of a Health Information System.

Course Ranking and Automated Suggestions through Web Mining

This paper introduces new metrics for course evaluation. It is also proposes a ranking algorithm that classifies courses based on the previous course evaluation metrics and suggests appropriate actions for course content improvement. The algorithm was tested and verified successfully in data originated from the eClass platform of TEI Kavala.

Greeklish: An experimental interface for automatic transliteration

“Transliteration” in linguistics means the system of conveying as nearly as possible by means of one set of letters or characters the pronunciation of the words in languages written and printed in a totally different script. This term may be applied to a transcription in Latin letters of Greek, Hebrew, or the Slavonic languages written in the Cyrillic alphabet. We present in this article Greeklish, a Windows application that automatically produces English to Greek transliteration and back-transliteration (retransliteration). This transliteration is based on an algorithm with a table of associations between the two character sets. This table can be modified by the user so that it can cover personal preferences or formal present and future rules. The novelty of this system is its speed of operation, its simplicity, and its ease of use. Our examples use a Greek to Latin (English) alphabet mapping, but the Greeklish application can easily use any X to Latin mapping, where X is any non-Latin alphabet.

Semantic Web Multimedia Metadata Retrieval: A Music Approach

The World Wide Web hosts ever-increasing volumes of diverse data and associated services that cannot be fully exploited unless appropriate management and integration technologies are developed. To cover this need, the Semantic Web was introduced as the second-generation WWW, enriched by machine-understandable data which supports the user in his tasks. The rapid growth of the Web has produced an explosion of the digital media availability, combining the media reproduction of the last decades, with easy and fast distribution. Therefore, an extensive use of metadata (the fundamental building block of the Semantic Web) for digital media is required. An approach to multimedia metadata processing is proposed, using rule reasoning based on RDF model theory, so that users can issue Semantic based query by interaction interface and achieve information sharing. RDF is widely used for data exchange and to publish data from database systems to the Web by providing input to content generators for Web pages. This paper describes a Semantic Web-based application using RDF to query an online music metadatabase. The RDF mapping of the retrieved multimedia metadata is customized to achieve approximate and trustworthy query answering and content management.

ALBL: an adaptive load balancing algorithm for distributed web systems

This paper presents an adaptive load balancing algorithm (ALBL) for cluster-based web systems. The balancing policy is based on two criteria: HTTP process time and network delay. The former describes web server ability to process a forthcoming request, while the latter tries to estimate network conditions. Periodic criteria calculations are performed by agents at the web switch and the weight estimation process is transparent to web servers enhancing therefore distributed systems scalability. We put to test our implementation against known blind selection balancing algorithms used at web-farms such as: Round Robin, stateful ones such as Least Connections and adaptive ones such as Least Load. We also put to test performance and scalability of previous algorithms. From our testbed scenario results we show that ALBL algorithm outperforms stateless and stateful algorithms and also presents significant performance gains towards adaptive algorithms. We also show that our algorithm scales well as the number of balancing servers and web clusters requests rate increases. We also pinpoint ALBL algorithm ability to predict network conditions and web servers load without the use of feedback information obtained by web servers but frequent service and network probes towards web servers issued by the web switch.

A Combined Algorithm for LMS Usage Assessment

LMS are used more and more nowadays. There are some algorithms for LMS usage assessment. The main problem is that these algorithms are rarely combined. This gap can be filled by using S-Algo which combines ranking and suggestion results of two or more suggestion/ranking algorithms into an ultimate and efficient ranking suggestion. Such efficient ranking is based on ranking algorithms used and proposed metrics recorded course attributes. S-algo was applied to Open eClass LMS tracking data of an academic institution. In the context of course assessment from student logged data, two existing algorithms called SUGAL and CCA as ranking/suggestion algorithms were selected. Then, on the results provided from these algorithms, S-algo tested and confirmed the success of its ranking process.

Snmp for Ethernet Networks SETH: A Network Benchmark Toolkit for Managing Routers Statistical Information

In this paper authors present a new open source SNMP manager application called SETH. This application includes the following major attributes over existing management tools: MIB file parsing capabilities, automatic discovery of network topology and variable up to real-time probing frequency of network equipment in order to extract useful statistical information regarding network conditions. Network information is acquired from network devices using real-time SNMP probes close to agent systems clock granularity. Then re-configurable network metrics are calculated and visualized. Additionally, SNMP trap or notification messages send by routers, indicating cases of network emergencies, are also recorded. Graphical visualization of metric values, network correlation functions and support for all SNMP protocol versions is also provided by the tool.

Balancing HTTP traffic using dynamically updated weights, an implementation approach

In this paper we present a load balancing application for HTTP traffic that uses dynamic weights. We introduce a load balancing policy based on two criteria: “process time” and “network delay”. The former describes Web servers ability to process a forthcoming request, while the latter tries to estimate network conditions. Calculation of the two criteria is periodically updated. A Weighted Round Robin algorithm was implemented using the two aforementioned metrics in order to dynamically estimate the balancing weights. We confirm that the combination of the two criteria increases sensitivity and responsiveness of the application towards network conditions and therefore the performance of the whole load balancing system. Balancing decisions should not be only “load” or “connection” dependent, but also contention dependent.