This action might not be possible to undo. Are you sure you want to continue?
6, June 2011
Additive model of reliability of biometric systems with exponential distribution of failure probability
Zoran Ćosić(Author)
director Statheros d.o.o. Kaštel Stari, Croatia zoran.cosic@statheros.hr Bihać, Bosnia and Hercegovina jacosic@gmail.com
Miroslav Bača (Author)
professor Faculty of Organisational and Informational science Varaždin, Croatia miroslav.baca@foi.hr
Jasmin Ćosić (Author)
IT Section of Police Administration Ministry of Interior of Unasana canton
Abstract— Approaches for reliability analysis of biometric systems are subject to a review of numerous scientific papers. Most of them consider issues of reliability of component software applications. System reliability, considering technical and software part, is of crucial importance for users and for manufacturers of biometric systems. In this paper, the authors developed a mathematical model to analyse the reliability of biometric systems, regarding the dependence of components with exponential distribution of failure probability. Keywords Additive model, Biometric system, reliability, exponential distribution, UML,
Data collection subsystem consists of a biometric sample, method of sampling, and sensors that are sampled. Signal processing subsystem consists of drainage structures, quality control and comparison of samples. Decision subsystem consists of the decision mechanisms and storage subsystems. Schematisation of model described in figure 1 can be shown on figure 2.
I.
INTRODUCTION
The general biometric system, according to [1] Wyman shown in Figure 1, consists of 5 elements which are located in all biometric systems today.
Figure 2 Schematic presentation of a biometric [1] system is a simplified representation of a system in Figure 1 and shows the serial configuration of system components dependence II. THE DEFINITION OF THE RELIABILITY OF BIOMETRIC
SYSTEMS
Figure 1. [1] Each subsystem consists of the elements that contribute to the overall system quality.
Biometric system designers and producers are motivated to use already constructed components and modules. Component system has a high reliability expectations , no matter who is producer. Most of existing reliability models are so generally described to not consider particularity of components and modules. In this paper authors will describe methodology based on mathematical model which take account of component reliability and connections reliability also. UML methodology in describing system and his inner interaction, simplify approach for researchers. UML [2] is also becoming standard in the process of designing and manufacturing systems so production of component systems gets benefits from the UML representation.
1
http://sites.google.com/site/ijcsis/ ISSN 19475500
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 9, No. 6, June 2011
Assessment of [3] generic biometric system reliability in UML is given in Figure 3, which describes the use of Use Case diagram:
Diagrams sequences play an important role in assessing the reliability of the system because they give information on how many components are involved in the execution of a scenario. Through sequence diagrams it is simple to count the periods of availability of components in the given scenarios as shown in figure (3). The probability of failure of components with known busy periods, can be given by the following expression: (3)
Figure 3 [3] q1 and q2 represent the probability that users u1 and u2 will access the system using some of its functionality. P11 and P12 represent the probability that user u1 will use the functionality of f1 and f2, and P21 and P22 represent the same probability for the user u2. The probability of execution of use case x, is defined by the expression: (1) m is number of users. If we are able to join a no uniform distribution to Diagram of sequences in a given use case then (1) can be expressed as: (2)
In witch is:  probability of component failure i in the scenario j  occupancy time of component i in the scenario j The expression (3) applies only if the following conditions are met:  Independence of failure: the probability of failure of one component does not depend on other components

Regularity of failure: the probability of failure of one component is equal throughout the execution of occupation period of the component
You can also show every moment of occupancy of any component of the system considering the method to be executed at that moment in the scenario If we replace where is with a set of method of failure probability then equation (3) becomes: (4) During system operation, components interact and exchange information. Then it is necessary to take into consideration occupation period of the component: (5) Where is θij the probability of failure of system components and bpbusy period of the system. (6) Where is Ψlmj the probability of failure connections between the components and  the number of interactions between system components. (7)
Where the fj (k)  frequency of the kth transition of sequence diagram in the jth case. P (kj) – presents a probability of default scenarios.
Figure 4 Sequence diagram [3]
The reliability of the system taking into account the probability of failure can be expressed as:
2
http://sites.google.com/site/ijcsis/ ISSN 19475500
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 9, No. 6, June 2011
n2( Δt )  number of failures in certain time interval Δt n1( t − Δt ) – the non failed number of elements at the end of
(7a)
the interval
Δt , or until t − Δt
(7b)
The intensity of the element failure is calculated by the expression:
III.
RELIABILITY PREDICTION OF THE COMPONENTS SERIAL
DEPENDENCE WITH EXPONENTIAL DISTRIBUTION
λEL
Where is:
)
1 1 =Θ= n Θ⋅n
(11)
Based on the above assumptions [4], [5] system reliability can be calculated according to the law of exponential distribution: (8) In wich are: Rs  System reliability λ  Intensity of system fault t  Required time of reliable operation of system Proof:
n number of usable parts of the confidence interval
(1 − α ) = 0, 75
Θ  lower limit of confidence for the mean time between
failures The total intensity of failure taking into consideration the number of elements that are not failed in a given time is calculated by formula:
(12) Where is: nEL number of elements of subsystems that are not failed In a serial dependence between of all parts of the system the failure of any part of the system may cause the failure of the entire system. Failure intensity function λs in the case of a serial dependence of system elements is calculated by the expression: (9) Where is λi  failure intensity of the ith part of the system. Failure intensity function λs is equal to the ratio between the number of failures in the timeframe and the correct number of elements in the system, until the beginning of this interval: Calculation of reliability [6] of some element is based on empirical data on the time of functioning and eventual failure of the element. Problem [7], [8] becomes more complex when the information about the failure doesn't exist. In case that part worked perfectly, and information about the exploitation are available. If one assumes that the given part can apply the rule of the exponential distribution, it is possible to determine the upper limit of confidence for the intensity of failure, in the cases of continuous operation or one or more failures. Lower limit of confidence for the mean time between ) failures Θ , for confidence interval (1 − α ) is calculated using the formula: Mean time between failures MTBFs can be calculated using the expression:
(13)
(10) Where is: λs function of failure intensity of system Δt  failure time of an system element
3
http://sites.google.com/site/ijcsis/ ISSN 19475500
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 9, No. 6, June 2011
) Θ≥
Where is:
χ 2α ,2 r + 2
2tr
=
χ 20.25,2
2tr
(14)
probability. In accordance with this authors will define mathematical model for recovery system probability than system readiness to use.
tr – total time of system operation r – Number of elements that have failed
REFERENCES
[1] [2] [3] Zasnivanje otvorene ontologije odabranih segmenata biometrijske znanosti  Markus Schatten– Magistarski rad – FOI 2007 Modelling biometric systems in UML – Miroslav Bača, Markus Schatten, Bernardo Golenja, JIOS 2007 FOI Varaždin A Bayesian Approach to Reliability Prediction and Assessment of Component Based Systems – KH. Singhy, V. Cortellessa, B. Cukic, E. Gunely,V.Bharadwaj Department of Statistics, Lane Department of Computer Science and Electrical Engineering West Virginia University, Proceedings of the 12th International Symposium on Software Reliability Engineering (ISSREí01) 10719458/01 2001 IEEE Simulacijsko modeliranje pouzdanosti tehničkog sustava brodskog kompresora – Zoran Ćosić – Magistarski rad  2007 Teorija pouzdanosti tehničkih sistema, Vujanović Nikola, Vojnoizdavački novinski centar, Beograd 2005 The Impact of Error Propagation on Software Reliability Analysis of Componentbased Systems – Petar Popic, Master thesis, West Vriginia 2005
χ 2α ,2r + 2
IV.
 Random variable which has distribution
SPECIAL CASE OF NOTFAILURE SYSTEM
[4] [5]
Considering (8), (10) and (14) we obtain an expression for the reliability of the whole system: (15) In futher calculations: (16) Taking into account (11) we obtain an expression for the reliability of the each elements of the system: (17) V. CONCLUSION AND FURTHER RESEARCH
[6] [7] [8]
The reliability of technical systems is the subject for many research scientists, according to that analysis are available in different models of reliability of biometric systems that in most cases take into account the software as one of the components of the same system. This study developed mathematical model of a generic biometric system that consider user, hardware and software influence and assumes serial dependence of the components and the exponential distribution of failure probability of system components. Scientific contribution is expressed through mathematical model definition of biometric system reliability prediction within special case where is not possible to get failure data or in the case of perfectly working system. The results of calculations can prevent real failure events by defining preventive maintenance. This mathematical model allows the prediction of system reliability in the early fase of projecting of its components . The subject of further research will be to create an integrated mathematical model of reliability of complex systems that considers the parallel and combined dependency of system components with different distributions of failure
AUTHORS PROFILE Zoran Ćosić, CEO at Statheros ltd, and business consultant in business process standardization field. He received BEng degree at Faculty of nautical science , Split (HR) in 1990, MSc degree at Faculty of nautical science , Split (HR) in 2007 , actually he is a PhD candidate at Faculty of informational and Organisational science Varaždin Croatia. He is a member of various professional societies and program committee members. He is author or coauthor more than 20 scientific and professional papers. His main fields of interest are: Informational security, biometrics and privacy, business process reingeenering, Jasmin Ćosić has received his BE (Economics) degree from University of Bihać, B&H in 1997. He completed his study in Information Technology field (dipl.ing.Information Technlogy) in Mostar, University of Džemal Bijedić, B&H. Currently he is PhD candidate in Faculty of Organization and Informatics in Varaždin, University of Zagreb, Croatia. He is working in Ministry of the Interior of Unasana canton, B&H. He is a ICT Expert Witness, and is a member of Association of Informatics of B&H, Member of IEEE and ACM. His areas of interests are Digital Forensic, Computer Crime, Information Security and DBM Systems. He has presented and published over 20 conference proceedings and journal articles in his research area Miroslav Bača is currently an Associate professor, University of Zagreb, Faculty of Organization and Informatics. He is a member of various professional societies and program committee members, and he is reviewer of several international journals and conferences. He is also the head of the Biometrics centre in Varaždin, Croatia. He is author or coauthor more than 70 scientific and professional papers and two books. His main research fields are computer forensics, biometrics and privacy professor at Faculty of informational and Organisational science Varaždin Croatia
4
http://sites.google.com/site/ijcsis/ ISSN 19475500