Abstract. My title was inspired by prof. Akira Imada's talk at ACS-CISIM 2006 ("When a Family of Iris Flower is Normal, then are Others Abnormal?").

Prof. Imada considered the problem in the context of Intrusion Detection.

My question is: when dealing with real data, may we expect that we will deal with a 'Gaussian' distribution?

It is believed widely, that the 'normal' (Gaussian) probability distribution is the most natural one, arising from a multitude of small independent contributions, which add to yield the final value of the observed variable. A great majority of statistical tests is based on the assumption, that the observed variable has a (multivariate) normal

distribution.

How often does it happen that real data have a (multivariate) normal distribution? It appears that this happens very rarely.

The departures from normality (Gaussianity) are caused very often by heterogeneity of the data.

In my presentation, I will look at the data with the aim of detecting non-homogeneity. Very often, real data may be modeled as a finite or infinite mixture composed from compound distributions - the compounds being Gaussian or Student's t distributions. This happens e.g., for financial data (stock returns) or internet traffic data. In particular, the importance of Student's t distribution will considered.

New Approach for Computer Support of Strategic Decision Making in Enterprises by means of new Class Understanding Based Management Support Systems

Ryszard Tadeusiewicz, Marek Ogiela, Lidia Ogiela

AGH University of Science and Technology, Krakow, Poland

{rtad, mogiela, logiela}@agh.edu.pl

Abstract. Automatic understanding is new term, which three or four years ago was almost invisible in bibliography, now is used by still biggest and biggest number of researchers and developed in increasing number of papers. Most of such papers take into account automatic understanding of the information in the traditional text form – sometimes on the basis of such ideas as Semantic Networks [1] and Ontologies [2]. Automatic understanding is then treated as s tool for better web searching and document summarization.

Semantic content in text is something evident, therefore every scientific paper with new ideas on automatic understanding of the merit of written documents is accepted and considered as a serious achievement on the way toward natural and easy communication between man and machines. In contrast to this situation when we started talking about automatic understanding of the images [3] almost everybody wrote papers against this idea, because semantic content of the image can be seen as something strange. Now it is evident, that image can have merit sense and the papers describing details of automatic understanding of the images are accepted by the Editorial Boards of the best scientific journals [4].

Idea of automatic understanding is not limited to the area of text messages and image presentations only. Important knowledge may be also hidden in other form data, for example in couples of numbers representing business data. Such data are collected, stored in databases, and used for analysis and presentations. Results of data analysis are used in decision making processes, because only precise and up-to-time information can lead to good decision and successful control. Therefore computerized economical information systems are used by everybody and everywhere.

Such economical information systems are build in most cases in form of specialized databases, sometimes more sophisticated data warehouses, powered by management information systems and also by (most fashionable) decision support systems.

All the systems mentioned above are very useful, but its application is limited to the lower or middle level of management. It is because typical decision support system can transform raw data into useful information, but cannot convert information to the wisdom. On the operational or tactic level of decision making processes information is enough, but when we approach to the strategic level of management – we do need wisdom, and until now such wisdom can be created only in human mind (even not every because for strategic planning a special kind of genius is necessary).

Sometimes we have genius on the top-level executive board, but often we have not. In such situation we need computer devices for aid the creation of wisdom on the basis of collections of data and files of information. If we remove emotions connected with the words “genius”, “strategic planning” and “top level management” we can discover, that everything what we need in fact can be encapsulated into automatic understanding module.

Automatic understanding of economical data for management and decision making purposes was proposed in some previously published papers (see [5], [6]), but in the presented paper will be described in details. In paper will be proved, that automatic understanding of the business situations and opportunities can be designed in very similar way as automatic understanding of the medial images, but in definitely different form than algorithms used for automatic understanding of the merit content of text messages based on ontologies. The details of the considered method will be presented in full text.

Bibliography:

1. G. Antoniou, F. van Harmelen, A Semantic Web Primer, The MIT Press, 2004.

2. A. L. Rector, J. E. Rogers, Ontological Issues in GALEN, Part I: Principles, Part II: High Level Schemata, in Methods of Information in Medicine, 2002.

3. Tadeusiewicz R., Ogiela M. R., Medical Image Understanding Technology, Series: Studies in Fuzziness and Soft Computing, Vol. 156, Springer-Verlag, Berlin – Heidelberg – New York, 2004

4. Ogiela M.R., Tadeusiewicz R., Ogiela L., Image Languages in Intelligent Radiological Palm Diagnostics, Pattern Recognition, Vol. 39/11. 2006, pp. 2157-2165

5. Tadeusiewicz R., Ogiela L., Ogiela M.R., Cognitive Analysis Techniques in Business Planning and Decision Support Systems, in Rutkowski L. (et al. eds.): Artificial Intelligence and Soft Computing, Lecture Notes in Artificial Intelligence, Vol. 4029, Springer-Verlag, Berlin – Heidelberg – New York, 2006, pp. 1025-1039

6. Tadeusiewicz R., Ogiela M.R., Why Automatic Understanding? In Beliczynski B. et al. (Eds.): ICANNGA 2007, Part II, Lecture Notes on Computer Science, vol. 4432, Springer-Verlag, Berlin – Heidelberg – New York, 2007, pp. 477 – 491

IEEE Distinguished Visitor Lecture

Agent Systems and their Applications – a Biased Perspective

Marcin Paprzycki

Computer Science Institute

Warsaw School of Social Psychology

Warsaw, Poland

http://mpaprzycki.swps.edu.pl

Systems research Institute

Polish Academy of Science

Warsaw, Poland

marcin.paprzycki@ibspan.waw.pl

Abstract

Since 1994 we are supposed to believe that software agents will become the next revolutionary information technology that is going to solve, among others, the problem of information overload [4]. This change will not only impact human-computer interaction [2, 4] but also change the way in which we construct software – especially in the case of large complex systems [3]. However, as it is easy to see, the agent-revolution does not seem to materialize. Even though, software agents are very popular subject of academic research, it is difficult to point to a successful real‑world implementation of a large scale agent system.

The aim of our presentation is two-fold. First, a few basic observations about software agents and agent systems will be presented. Second, it will be shown that it is possible use modern agent platforms (e.g. JADE) to implement large scale agent systems [1]. Finally, sample applications of software agents will be briefly discussed (based on [5]).

References

[1] K. Chmiel, M. Gawinecki, P. Kaczmarek, M. Szymczak, M. Paprzycki (2005) Efficiency of JADE Agent Platform. In: Scientific Programming, 13(2), 2005, 159-172

[2] J. Hendler (1999) Is There an Intelligent Agent in Your Future?, Nature, 11 March, 1999

[3] Jennings N. R. (2001) An agent-based approach for building complex software systems, CACM, 44 (4), 35-41

[4] Maes P. (1994) Agents that Reduce Work and Information Overload, CACM, 37(7), 31-40

[5] Papers completed by our team can be found at: http://agentlab.swps.edu.pl

Quantifiable Quality of Service includes
information security

Svein Knapskog

Department of Telematics,
Faculty of Information Technology,
Mathematics and Electrical Engineering,

The Norwegian University of Science and Technology (NTNU)

Abstract

Information Security as part of Quantitative Quality of Service may be represented by modeling, design, analysis and evaluation of information security for ICT systems, services and protocols, has in the past most frequently been achieved by using qualitative principles, methods and measures. However, in the present transition into an end user environment dominated by ubiquitous computing power and service realization, the information security research needs to be focused on the quantitative security aspects of description and negotiation of perceived QoS.

Present development is in the direction of an ever increasing number of mobile units connected by wireless networks with a highly dynamic architecture. The development of new services with previously unknown or unexploited functionality is often prioritized above security, and it is therefore quite challenging to educate developers to focus on security as a pervasive capability for the whole lifespan of a product or a service. One important general objective is to consider information security an integrated part of the quality of service (QoS), and have security expressed in quantitative terms. Methods for development and evaluation from the reliability area may be adapted so to be applicable also for quantitative security evaluation. The information security research at Q2S currently focuses on more specific issues as well, such as cryptographic protocols and use of game theory in modeling. To model human behavior is a major challenge, and the first fledgling attempts to apply game theory for the purpose have been made, with apparent success. It seems like game theory may have considerable potential when it comes to describing and simulating attacker behavior in a networked environment. Methods for dynamic assessment of risk connected to the use of a specific service will be developed, with the objective to make it feasible for the user to realistically judge whether the benefit of the service outweighs the risk incurred

Cryptographic protocols can be considered as distributed algorithms that employ cryptographic functions for securing distributed computerized communication systems. Cryptographic services, e.g. services providing confidentiality, authenticity, integrity and non-repudiation are services implemented by cryptographic mechanisms. Such services are necessary in the future open, distributed service provision environment of a networked world. Parts of the network may be ad hoc, with unreliable and untrustworthy nodes, and without any central control facility which could act as a trusted third party. A node in an ad hoc network will typically be restricted in its information processing and communication capabilities, making it even more difficult to deploy extensive security services for the users. This user environment enhances the uncertainty, even above the present level of wireless network environments. Our ongoing and future research will focus at increasing the understanding of which cryptographic services will be appropriate in such complex and diverse networked environments, how the security capabilities can be defined and modeled, and on analysis of the security of the protocols to be employed in the implementation of the cryptographic services.

A couple of examples of the types of research projects that have been performed will be shown.

"GUI Patterns and Web Semantics"

Václav Snášel

Department of Computer Science

VSB-Technical University of Ostrava
Czech Republic

Abstract

In this lecture we introduce a new approach to semantic analysis of web pages. To prove this approach we designed method for analysis and evaluation of web pages. The method is built on a silent agreement between web designers and users. The key aspects of this agreement are web patterns which are used by web designers in their web page implementations. With our method we can find out whether the pattern is presented on the page with high level of relevance. Key feature of our method is an independence on the page HTML code. In this lecture we explain essentials of our approach as well as key features of our method and context for proper usage. We present experiments which prove efficiency of the method and usefulness of the approach.

This lecture will be based on

M. Kudělka, V. Snášel, Eyas El-Qawasmeh, O. Lehečka, J. Tesařík: Domain Patterns and Semantic Annotation of Web Pages, the First IEEE International Conference on Digital Information Management, IEEE CS press, Bangalore 2006, India.

M. Kudělka, V. Snášel, Eyas El-Qawasmeh, O. Lehečka: Internet Searching Using Web Patterns, IEEE CS press, Dubai, IIT 2006.

M. Kudělka, V. Snášel, Eyas El-Qawasmeh, O. Lehečka: Semantic Annotation of Web Pages Using Web Patterns, IEEE/WIC/ACM WI-2006, Hong Kong.

M. Kudělka, V. Snášel, Eyas El-Qawasmeh, O. Lehečka, J. Pokorný: Semantic Annotation of Web Pages Using Web Patterns SITIS 2006, IEEE/ACM Springer Verlag, Tunisia, in print.

"Finding a needle in a haystack --- from the Baldwin Effect to

Finding a needle in a haystack --- from
the Baldwin Effect to Quantum Computation

Akira Imada

Brest State Technical University

Brest, Belarus

If we want to break someone else's PIN -- Personal Identification Number -- of, say, an ATM -- Automated Teller Machine --, how many trial would be necessary when we want to be efficient? This is a sort of what we call a-needle-in-a-hay-stack problem. In 1987, in their seminal paper, Hinton & Nowlan proposed a Genetic Algorithm with a needle being a unique configuration of 20-bit binary string while all other configurations being haystack. What they proposed was to exploit the Baldwin effect calling it a life-time-learning of individuals of their Genetic Algorithm. Since then there has been a fair amount of exploration of this problem, claiming, "This is a-needle-in-a-hay-stack problem, and we've found a more efficient algorithm than a random search." Some of them, however, were found to be the results of an effect of like-to-hear-what-we-would-like-to-hear. In this talk, we will try a bird's eye view on what we have as examples on this issue and how they were explored, including the approach by means of Quantum Computation which claims, "The steps to find a needle are order square-root of N while those of exhaustive search are order of N where N is the number of search points."

EYE REGION IMAGE FEATURE EXTRACTION FOR APPLICATION OF BIOMETRIC PERSONAL IDENTIFICATION

Ryszard S. Choraś

UTP, Bydgoszcz

Biometrics is defined as the measurable physiological and/or behavioral characteristics used to verify the true identity of an individual. Among a the physical features measured are the face, fingerprints, palm prints, hand or palmar geometry, retina and iris and vein. The behavioral features are signature or handwriting, keystroke pattems, gait motion, gesture and voice in a behavioral aspect.

Essentially a biometric system follows two characteristic traits: identification and verification. The former involves identifying a person from all biometric measurements collected in a database. The question that this process seeks to answer is: “Do I know who you are?” It, therefore, involves a one-compared-to-many match, also referred to as a ‘cold search’. Verification involves authenticating a person’s claimed identity from his/her previously enrolled pattern. “Are you who you claim to be?” is the question that this process seeks to answer. This involves a one-to-one match.

Various humans region biometrics have been used extensively in the areas of recognition and authentication. The biometric of eye region hold a key to identifying and classifying particular affective and cognitive states.

In this paper, a touch-free non-intrusive technique of extracting personal eye biometric data is proposed and analyzed.

The eye region contains many potential features of interest to the biometric researcher. The specific biometrics central to our effort are the irises, pupils, eyelids, eye folds, eyebrows, blink characteristics.

We detail describe the eye region based biometric method includes:

· iris of a human eye recognition (analyze features found in the colored ring of tissue that surrounds the pupil),

· retinal vessel segmentation and recognition,

· blinking - biometric indicator measures how the eye region changes as individuals blink specific patterns,

· geometrical features of eye region.

An Information-Based Approach to Cryptograph

Robert E. Hiromoto

Department of Computer Science

University of Ihado

USA

Modern techniques for cryptanalysis are based on Claude Shannon’s information theory. Shannon laid down the mathematical foundations for analyzing the information content of electronic communication, and extended that work to cryptography and cryptanalysis. Cryptography achieves security by using non-linear techniques to obscure the plain text content of messages. Techniques such as confusion and diffusion are employed in algorithms that exhibit a level of computationally ease to apply; yet, results in an encryption difficult to reverse. In today’s world of high-speed processors and large core memories, Shannon’s work on cryptography was clearly limited by the computational resources available to him.

In this presentation, we outline an approach that extends Shannon’s work by the introduction of information as a priori property sets that more accurately characterizes the encrypted languages and process. We apply the Set Membership Theory (SMT) that

provides a convenient framework to categorize and apply language statistics, and in particular treat the problem within the approach of Set Theoretic Estimation (STE). The STE approach has been applied to decrypt several simpler cipher techniques, including the shift and substitution ciphers. In this presentation, STE is applied to a more difficult block cipher technique. The results indicate that the STE approach is promising and represents the first application of the Set Theoretic Estimation (STE) technique to cryptography and cryptanalysis.

Toward Higher Level Intelligent Systems

Kurosh Madani

Images, Signals and Intelligence Systems Laboratory (LISSI / EA 3956)

PARIS XII University, Senart-Fontainebleau Institute of Technology,

Bât.A, Av. Pierre Point, F-77127 Lieusaint, France,

madani@univ-paris12.fr

Much is still unknown about how the brain trains and self-organizes itself to process so complex information. However, the recent advances in “neurobiology” allowed highlighting a number of animal’s intelligence chief mechanisms. In fact, our simple and inappropriate binary technology remains too primitive to reproduce the biological complexity of these marvels mechanisms, but a number of those highlighted points could already be sources of inspiration for higher level intelligent artificial systems. Among appealing features of animal’s brain, one can emphasize its “modular” architecture, its “self-organizing” capabilities and its “self-evaluation” ability. If it is still early to state on “concurrent” or “cooperative” nature of ways that these complex features interact, they could already be considered as basic features in emergence of higher level artificial intelligent behavior.

On the other hand, overcoming limitations of conventional approaches thank to their learning and generalization capabilities, Artificial Neural Networks (ANN) made appear a number of expectations to design “intelligent” information processing systems. If learning and generalization capabilities of these bio-inspired connectionist models appear as central requirements in intelligent systems’ design, nowadays, it is well admitted that intelligent behavior requires more sophisticated mechanisms than those performed by these “simple” models.

The main goal of this talk and the related paper are to show how these primary supplies could be exploited and combined in the frame of “soft-computing” issued techniques in order to design intelligent artificial systems emerging higher level intelligent behavior than those of conventional Artificial Neural Networks (ANN) based structures. These foremost features have inspired a set of implementations dealing with real-world applications covering several different areas as: robotics, image processing and pattern recognition, classification and dynamic nonlinear behavior modeling (identification and prediction).

Introduction of Humatronics - Towards Integration of Web Intelligence and Robotics

Yasufumi Takama

Tokyo Metropolitan University

Japan

In recent years, various systems based on electronics, such as cars, computers and robots have been getting into our daily lives. Unfortunately, asymmetrical relationship exists between human and such systems, imposing burdens especially on elderly or handicapped persons. In order to solve various problems for balancing such asymmetrical relationship, humatronics is proposed. Main purpose of humatronics is to establish the symmetric interaction between human and electronic systems by giving the systems the capability of understanding humans. Another important factor for humatronics is computer networks, over which the systems can share knowledge, information, and experiences. This lecture introduces the concept of humatronics and key technologies for realizing it. A prototype system that is currently developed based on the concept of humatronics is also introduced.

Benchmark problems in signal processing and control for testing emergent architectures and programming techniques of DSP's, FASIC's, and PLD's

Adam Dąbrowski and Paweł Pawłowski

Poznan University of Technology

Dept. of Computing & Management

ul. Piotrowo 3A

60-965 Poznan, POLAND

Commercially available digital programmable devices intended for intensive real-time computations, namely microcontrollers, DSP's, FASIC's, and PLD's, are typically developed to realize commonly accepted benchmark tasks such as solving difference equations, computing signal correlations, coding of audio and/or video streams, filtering of signals in the primary domain (mainly using FIR filters or filter banks based on the discrete wavelet transformation) or in the transform domain (e.g., by means of the FFT or DCT), etc. They, however, exhibit many disadvantages in various other modern demanding applications such as e.g. data acquisition of fast phenomena in metrology systems, control and measurement of vehicles, medical diagnostic and therapeutic systems and in many others, which require effective acquisition of large and irregular streams of data, real-time generation of precise signals, special techniques for data transmission, processing, and acquisition as well as scalability of functionality and built-in self-testing or even self-repairing facilities. That is why we present some carefully selected applications, which, as we believe, can serve as important additional benchmarks for a wide class of the applications mentioned above. Among the proposed problems are generators of various precise and quite sophisticated signals and microcontroller based data acquisition and measurement systems.

Due to the proposed multi-criteria selection of the benchmark problems we took both typical and somehow untypical tasks into account. In result we suggested four additional benchmark applications: measurement of vehicle crashes, intended for testing road restraint systems, generation of precise tunable signals for advanced audiometric tests, generation of precise photo-stimulation signals, and generation of precise multiphase sequences for control of discrete-time (e.g., switched capacitor) filters.

Nature Inspired Intrusion Detection Programs

Ajith Abraham

School of Computer Science,
Yonsei University,
134 Shinchon-dong, Sudaemoon-ku,
Seoul 120-749, Republic of Korea

Faculty of Information Technology,Mathematics and Electrical Engineering,
The Norwegian University of Science and Technology (NTNU)

Email: ajith.abraham@ieee.org
WWW: http://www.softcomputing.net

Abstract: Several information security techniques are available today to protect information systems against unauthorized use, duplication, alteration, destruction and virus attacks. An intrusion detection program analyzes what happens or has happened during an execution and tries to find indications that the computer has been misused. In this talk, we present some of the challenges in designing efficient intrusion detection systems using nature inspired computation techniques which could provide high accuracy, low false alarm rate and reduced number of features. Finally, we present some future research directions and the development of intrusion prevention systems.

References:

Ajith Abraham, Crina Grosan and Carlos Martin-Vide, Evolutionary Design of Intrusion Detection Programs, International Journal of Network Security, Vol.4, No.3, pp. 328–339, 2007.

Ajith Abraham, Ravi Jain, Johnson Thomas and Sang Yong Han, D-SCIDS: Distributed Soft Computing Intrusion Detection Systems, Journal of Network and Computer Applications, Elsevier Science, Volume 30, Issue 1, pp. 81-98, 2007.

http://top25.sciencedirect.com/index.php?cat_id=10&subject_area_id=7&journal_id=10848045