CC BY
28
R&RATA # 3 (Vol.1) 2008, September
MATHEMATICAL MODELS AND SOFTWARE TOOLS FOR QUALITY AND RISK MANAGEMENT ACCORDING STANDARD REQUIREMENTS
Prof. Andrey Kostogryzov
Director of the Research Institute of Applied Mathematics and Certification, 1-st Miasnikovskaja Ul., Vl. 3, Moscow, Russia,107564, tel. +7(495) 795-8524, e-mail: [email protected]
Abstract
The offered mathematical models and supporting them software tools complexes (M&STC) are purposed for a systems analysts from customers, designers, developers, users, experts of testing laboratories and certification bodies, as well as a staff of quality maintenance for any complex system etc. M&STC are focused on providing system standard requirements on the base of modeling random processes that exist for the life cycle of any complex system. Models implement original author's mathematical methodology based on probability theory, theory for regenerating processes and methods for system analysis. M&STC may be also used in training and education for specializations "System engineering", "Software engineering", "System safety and security", "Information systems".
1. Introduction
According to standard ISO/IEC 15288 system is defined as a combination of interacting elements organized to achieve one or more stated purposes. An application of offered methodology uses to evaluate probabilities of "success", cost, time and quality risks and related profitability and expenses. This helps to solve on the scientific basis the next practical problems in system life cycle: analysis of quality management systems for enterprises, substantiation of quantitative system requirements to hardware, software, users, staff, technologies; requirements analysis, the evaluation of project engineering decisions; investigation of problems concerning potential threats to system operation including information security and protection against terrorists; evaluation of system operation quality, substantiation of recommendations for rational system use and optimization etc.
2. Focusing on rational management
All complexes are offered for providing rational management. Management is a purposeful changing of an object state, a process or a system. Management is based on choosing one among a set of alternatives. Rational management is a management leading to the objective achievement according to the criterion of a chosen parameter extreme (minimum or maximum) under the set limitations. Classical examples of rational management are usually either maximization of a profit (an income, a degree of quality or security, etc.) under limitations on expenses or expenses minimization under limitations on an admissible quality and/or security level. It is clear that criterion and limitations may vary throughout the system life.
For rational management of processes it is necessary to know and plan their behaviour at various influences. For this purpose we offer for using about 100 the mathematical models [1,2]. As criterion parameters there are used the quantity measures (objective functions) characterizing a possibility of object achievement at different stages of a system life cycle. For example, an investor's criterion is the maximum income from the project implementation under limitations on
the production process and product quality. For the enterprise it is important to organize a quality management system properly - so that in the form of criterion it can choose the probability of qualified work performance, i.e. in time and without defects or the maximum probability of success for quality management policy concerning work complexes. A security service must provide safety of an object, a process or a system up to the mark. In this case there may be used the criterion of expenses minimum under limitations on the admissible level of dangerous influence risk taking countermeasures into account or minimum of a dangerous influence risk at limitations on expenses. The customer and the developer are interested in the final result - in this case as an integrated parameter there may be used such criterion as the maximum part of functional operations carried out with the admissible quality or the relative degree of customer satisfaction with the limitations on quality or expenses.
The first from the offered models is Complex for Evaluation of Information Systems Operation Quality (CEISOQ) [1-2,5-6]. The development of CEISOQ was based on the general purpose for all information systems (see Fig. 1 and 2).
Fig.1 The purpose of information system in a SYSTEM
Quality of information producing
Fig. 2 The main CEISOQ window for choosing the model
The problems connected with usual computation of parameters are called direct operation research problems or analysis problems. The problems directed on a choice of variants maximizing
or minimizing values of objective functions under limitations are called inverse operation research problems or synthesis problems. The offered models and supporting software tools allow to solve both direct and inverse operation research problems.
3. Abstract formalization
This formalization is used to building a probabilistic space (Q, B, P), where:
Q-is a limited space of elementary events;
B - a class of all subspace of Q-space, satisfied to the properties of o-algebra [3];
P - a probability measure on a space of elementary events Q.
Because, Q={ak} is limited, there is enough to establish a reflection <ak^pk =P(®k) like that
pk>0 and ^ pk = l. Such space (Q, B, P) is built by the limited theorems for regenerative processes
k
[3-4] and also by using principal propositions of probability theory and well famous results for single and multi-units queuing systems. This probabilistic space (Q, B, P) is the essence of mathematical models to support an assessment of standard system processes.
4. Example of created mathematical model
Nowadays at system development and utilization an essential part of funds is spent on providing system protection from various dangerous influences able to violate system integrity. Under system integrity it means such system state when system purposes are achieved with the required quality under specified conditions of use. Such examples of dangerous influences are terrorists attacks, viruses or 'violators' influences, software defects events etc. As this problem wasn't studied carefully dangerous influences often reach their aims.
There are examined three typical technologies of providing protection from dangerous influences. In this paper it is illustrated only technology l that is based on preventive diagnostics of system integrity. Diagnostics is carried out periodically. It is assumed that except diagnostics means there are also included means of necessary integrity recovery after revealing of danger sources penetration into a system or consequences of negative influences. Integrity violations detecting is possible only as a result of diagnostics, after which system recovery is started. Dangerous influences on a system are acted step-by step: at first a danger source penetrates into a system and then after its activation begins to influence. System integrity is considered to be violated only after a danger source has influenced on a system. If to compare a system with a man technology l reminds a periodical diagnostics of a man's health state. If diagnostics results have revealed symptoms of health worsening a man is cured (and integrity is considered as recovered). Between diagnostics an infection penetrated into a man's body brings a man into an unhealthy state (a dangerous influence is realized and integrity is violated)-see Fig. 3.
Cases: 1 2 3 4 5
^^^ - tine between the ragtixrirg dagrcslics;
- a requredpadod TTq cfpermarert secire cpeiiiicn;
- as nirim^ there is two diagnostics during a recpred period T^ (the iHiElralicn cf T^ nidde);
- a required pericd has ended after the last dagrcstic;
- adaricer soulcehasperelвtedbefae1lhenext ciagr^cslic;;
- a dareer scurce has rct penэlIatedirIto system;
- a peretrated darder soice has activated befcre the rext dagrcstic;
- a p^^tated cbncer scuce has rrct activatedbefcre the rext dagnoslic
Fig. 3 The illustraticr cf system rescurces prctecticr agairst dargerous irfluerces by techrclcgy 1
The availability cf mears cf darger scurces tctal-lct detectirg ard existerce cf ways cf vtolated system mtegrity tctal-lct reccvery is cbligatory requiremert. The cffered mcdels are suppcrted by scftware tods CEISOQ.
Te system prctecticr from dargerous irfluerces may be evaluated if the rext characteristics are krcwr: frequercy cf irfluerces for peretratmg a darger scurce mto a system (a) ; mear activatid time cf a peretrated darger scurce (P); time betweer the erd cf diagrcstic ard the begmmrg cf the rext me (Tbetw); diagrcstic time ircludirg the time cf system mtegrity reccvery (T diag ); a required pericd cf system cperaticr (T req.) for irvestigaticr. There are pcssible the rext variams:
variart 1 - the assigred pericd Treq is less thar established pericd betweer reighbcrirg
diagrcstics (Treq < Tbetw.+Tdiag)
variart 2 - the assigred pericd Treq is mcre thar cr equals to established pericd betweer
reighbcrirg diagrcstics (Treq > Tbetw.+Tdiag).
Statemem 1. Urder the ccrditicr cf mdeperderce cf ccrsidered characteristics the probability cf dargerous mfluerce abserce fcr variart 1 is equal to
Pirfl.(1)(Treq) = 1 - Bperetr* Bactiv(Treq), (1)
where * - is convolution sigr, Bperetr(t) is the prcbability distributicr furcttor (PDF) cf time betweer reighbcrirg irfluerces for peretratmg a darger scurce, Bactiv(t) is the PDF cf activatid time cf a peretrated darger scurce, fcr mcdelmg Bperetr.(t)=1-e-at, Bactiv(t)=1-e-t/p.
Note. This formula (1) is used also for the evaluation of security for system operation without diagnostics. There is supposed that before the beginning of period Treq system integrity is provided.
Statemem 2. Urder the ccrditicr cf mdeperderce fcr ccrsidered characteristics the prcbability cf dargerous mfluerce abserce fcr variart 2 is equal to
Pinfl.(2)=Pmdl+Pend (2)
where Pmdi - is the prcbability cf dargerous mfluerce abserce withm the pericd Treq smce begmmrg to the last diagrcstic, Perd - is the prcbability cf dargerous mfluerce abserce withm the pericd Treq after the last diagrcstic, i.e. m the last remamder Trmr=Treq -N(Tbetw+ Tdiag), N is the umber cf pertods betweer diagrcstics placed whdly withm assigred pericd Treq,
N = [Treq/( Tbetw + Tdiag)] - is mteger part;
Pwhclly(i) =Pirfl.(i)(Tbetw +Tdiag), is calculated by formula (1),
N (Tbetw. + T diagN T rmn s v
Pmdl - Pwholly(1) , Pend - Pinf l.(1) (Trmn)* (3)
T req. T req.
The mathematical proof is in [1,2]. This is one from more than 100 mathematical models offered to support an assessment of standard system processes.
5. Software tools to support an assessment of standard system processes
35 created software tools complexes implementing original mathematical models [1-2,5-6] consist complexes created in 2001-2005: for the Evaluation of Information Systems Operations Quality (CEISOQ, CEISOQ+); for Evaluation of System Vulnerability including Conditions of Terrorist Threats ("VULNERABILITY"); for complex analysis of system security ("ANALYSIS OF SYSTEM SECURITY"), for Modeling of System Life Cycle Processes "MODELING OF PROCESSES". The last complex "MODELING OF PROCESSES" includes multi-functional complexes for evaluation of Agreement (models and software tools "ACQUISITION", "SUPPLY"), Enterprise (models and software tools "ENVIRONMENT MANAGEMENT", "INVESTMENT MANAGEMENT", "LIFE CYCLE MANAGEMENT", "RESOURCE MANAGEMENT", "QUALITY MANAGEMENT") and Project (models and software tools "PROJECT PLANNING", "PROJECT ASSESSMENT", "PROJECT CONTROL", "DECISIONMAKING", "RISK MANAGEMENT", "CONFIGURATION MANAGEMENT", "INFORMATION MANAGEMENT") Processes Modeling and also for Technical Processes Modeling (models and software tools "REQUIREMENTS DEFINITION", "REQUIREMENTS ANALYSIS", "ARCHITECTURAL DESIGN", "HUMAN FACTOR", "IMPLEMENTATION", "INTEGRATION", "VERIFICATION", "TRANSITION", "VALIDATION", "OPERATION", "MAINTENANCE", "DISPOSAL").
The models created have undergone extensive testing in an operational environment and the results have been compared with the results of other independent models (if such exist). This comparison has provided documented evidence that the models implemented in these tool suites are realistic, including the reality of the calculations and the time&probabilistic characteristics. Created software tools are an original Russian creation patented by Rospatent, certified, have been presented at seminars, working groups, symposiums, conferences and forums since 2000 in Russia, Australia , the USA, Canada, France, Germany, Kuwait. In 2001 the CEISOQ [1-2,5-6] was awarded be the Golden Medal of the International Innovation and Investment Salon, in 2004-2005 the software tools "RISK MANAGEMENT", "HUMAN FACTOR" and "ARCHITECTURAL DESIGN" also were awarded by the Golden Medal of the International Exhibition "Intellectual Robots" and acknowledged as the software products of the year.
How these models adequacy may be conformed? Though any answer to these questions won't be irrefragable for a certain system we shall try to formulate our arguments.
Argument 1. The M&STC uses mathematical models formalizing standard processes on time line. Majority of dependencies gives upper and lower estimations. The fact is that while shaping models all mathematical results are initially drawn in the integral form. As input data are somehow connected with time after choosing distribution functions characterizing these data there were selected the gamma - distribution and the Erlang's distribution. Mathematicians know that these distributions approximate sums of positively distributed random variables well. Every temporary data are as a matter of fact such a sum of compound time expenses. Studies of regularities have shown that extremes are achieved on bounds of these distributions, i.e. of exponential and deterministic (discrete) distributions. Thus, real values will be somewhere between lower and upper estimations calculated by the software tools.
Argument 2. As a basis of models there is used the probability theory and the theory of regenerative processes. Proofs of basic theoretical results are cited in [1-2]. If to return in the 70-s of the last century we may remember the boom of mathematical modeling, defining calls flow reliable and time-probabilistic characteristics. The boom passed and appeared the reliability theory,
the queuing theory and a variety of models, which proved themselves to be effective. There are created standards and other normative documents regulating system methodical evaluations on the basis of these models. Nowadays these models are widely used and trusted because they produce reliable results confirmed in the course of time. It is worth to remind that these created theories and models are based on the probability theory and the theory of regenerative processes. The several offered models "The model of functions performance by a system in conditions of unreliability of its components", "The models complex of calls processing", "The model of entering into system data current concerning new objects of application domain" are the classical adapted models of the 80-s improved to meet the requirements of the present time. The other models are created on the basis of the limit theorem for regenerative processes developed in the 70-80-s in Moscow State University on the faculty of computing mathematics and cybernetics by professor Klimov's school [4]. Three-year testing of M&STC including beta-testing by fifty different companies raise confidence in models algorithmic correctness.
Argument 3. Skilled analysts know that if a probabilistic analytical model is incorrect then if input data are changed in the range from -<x> to there are always errors appearing either in infraction the probability theory laws or in illogic of dependencies behavior (most probably on the bounds of possible values) or in impossibility of obtained effects physical explanation. Bounds of input data in the M&STC are assigned in the range from -<x> to or to be more exact from 10-8 milliseconds to 108 years.
Argument 4. As far as possible any designer tends to use several models of different authors. If results of different models use are not divergent a designer begins to trust not only to results but also to the models. Comparison of results of the M&STC use with results of other models use proved its high adequacy (concerning computations of reliability and time-probabilistic characteristics, the other models don't have analogues).
6. Examples of software tools application
The offered M&STC have been and are applied for solving the problems of:
information security and reliability for banks, transport systems, protected and military objects etc.;
rational protection for oil and gas systems in conditions of terrorist threats;
quality and reliability for cosmic robot systems and heat supply etc.;
risk analysis for dangerous coal mine and manufactures;
system certification;
education in the field of system analysis.
Below there are demonstrated some capabilities of the software tools "RISK MANAGEMENT".
Example 1 for demonstration the capabilities of subsystem "Evaluation of counteraction measures effectiveness". Let 10 barriers be installed in order to protect valuable resources of a system from unauthorized access. In table there are shown prospective characteristics of barriers (as counteraction measures) and the mean time of their possible overcoming by a specially prepared violator (as the time of keeping measure effectiveness). Real values of similar characteristics may be received as a result of natural experiments or application of other special models.
It is required to evaluate the risk of dangerous influence on a system in spite of counteraction measures during a week. The minimal admissible risk shouldn't be more than 0.0001. The initial data for calculations are shown in the Table.
Table
Characteristics of the threat scenario and the protection system
Barrier Change frequency of the barrier parameter value as time to the next strenthening of the measure Mean time of barrier overcoming by a violator Possible way of barrier overcoming
1. External guards Change of guards every 24 hours 10 hours Latent penetration
2. System of passes to the system with a change of security services Change of guards every 24 hours 10 minutes Documents falsification, conspiracy, fraud
3.The electronic key to get to the control unit 5 years (time between changes) 1 week Theft, forcible key withdrawal, conspiracy
4. The password to enter the automated system 1 month 10 days Spying, compulsory questioning, conspiracy, selection of a password
5. The password to get access to software devices 1 month 10 days —II—
6. The password to get access to the required information 1 month. 10 days —II—
7. The registered external information carrier with write access 1 year 24 hours Theft, forced registration, conspiracy
8. Confirmation of a user identity, during a session of work with the computer 1 month. 24 hours Spying, compulsory questioning, conspiracy
9. Telemonitoring Time between changes of software devices - 5 years 1 month Simulation of a failure, false films, dressing up as employees, conspiracy
10. Encoding of the most important information Change of keys every month 1 year Decoding, conspiracy
Solution. The analysis of the withdrawn calculated dependences has shown the following (see
Fig. 4).
The first 3 barriers as counteraction measures are overcome with the probability about 0.34. Use of alternating passwords once a month for the 4th, 5th and 6th barriers allows to decrease the risk from 0.34 to 0.14. However, the general system protection after the introduction of the first six barriers remains rather weak. The 7th and 8th counteraction measures are practically useless. Use of telemonitoring means allows to decrease risk of dangerous influence on a system in spite of counteraction measures to 0.002 what also doesn't meet the stated requirements. The use of all 10 counteraction measures provides the required system protection.
|i|i'I.Pmi).|ii)JIIH|ii.Hll.i.l II.. .1 HI . I I.. JSIJLI
m
□ ■ H
0.JJ1Í' H
ft AJI il H ftWOI H
ill I
ftUHl H í'f Tí'« H
Fig. 4 Results of computation for example 1
The use cf subsystem "Evaluattor cf experses for risk reterticr" allcws to defme experses agamst risks (see calculattor results d Fig. 5). It is the capability to cptimizmg by criteria "risk-experses".
Fig. 5 Results of computation by the subsystem "Evaluation of expenses for risk retention"
The use cf subsystem "Substartiaticr cf ccurteracticr strategy agamst risks" allcws to evaluate differem damages ard experses agamst risks for the giver scerarto.
The last ccmplex "ANALYSIS OF SYSTEM SECURITY" allcws to evaluate the mtegral security for the system ccrsistmg cf ary umber cf ccmpcrerts. The ccrditicr is the ccmpcrerts are umted ir parallel ard/cr ccrsecutive crder. The structure may be ary degree cf ccmplexity. The strergth cf every measure (ccmpcrert) is apprcximated by expcrertial lcw.
Example 2 _ for demonstration the capabilities of subsystem "Analysis of integral security". Let the system contains 3 level of interacted subsystems: higher subsystem, interim subsystem and subordinate subsystem (see Fig.1). It may be territorially distributed enterprise or bank with the branches etc. Every subsystem have the valuable resources protected by the counteraction measures from the table (see example 1). The system structure, constructed by software tools for modeling, is on Fig. 6. The frequency of threats source appearance is 10 times in a year by qualified violator, the mean time for system recovery is 1 hour.
Fig. 6 System structure for modeling
It is required to evaluate the probability of providing system security in a year for the given scenario of threats.
Solution. The integral analysis of the withdrawn calculated dependences has shown the following (see Fig. 7). The probability of providing system security without control and monitoring is about 0.19, with periodic control (without permanent monitoring for the 9-th and 10-th barriers) - 0.39. The use of all 10 counteraction measures (including permanent monitoring for the 9-th and 10-th barriers) provides system security with the probability more than 0.95 against general expenses 104000 conditional units .
Fig. 7 Results of computation by the subsystem "Analysis of integral security"
7. Conclusion
Expected pragmatic effect from application is the next. It is possible to provide essential system quality and security rise and/or avoid wasted expenses in system life cycle on the base of processes modeling by the offered mathematical models and software tools complexes.
8. References
[1] M.M. Bezkorovainy, A.I. Kostogryzov, V.M. Lvov Modeling Software Complex for Evaluation of Information System Operation Quality CEISOQ. 150 problems of analysis and synthesis and examples for their solutions, Moscow: Armament.Policy.Conversion, 2001, 222p.
[2] A.I. Kostogryzov, G.A. Nistratov Standardization, mathematical modeling, rational management and certification in the field of system and software engineering, Moscow: Armament.Policy.Conversion, 2004, 395p. (in Russian)
[3] W. Feller An Introduction to Probability Theory and Its Applications, Vol. II, Willy,
1971.
[4] G.P.Klimov Probability theory and mathematical statistics. Moscow State University, Moscow, 1983 (in Russian), 328p.
[5] A.I. Kostogryzov "Software Tools Complex for Evaluation of Information Systems Operation Quality (CEISOQ)". Proceedings of the 34-th Annual Event (25-29 September 2000) of the Government Electronics and Information Association (GEIA), 2000 Engineering and Technical Management Symposium, 2000, 63-70.
[6] A.I. Kostogryzov "Modeling Software Tools Complex for Evaluation of Information Systems Operation Quality (CEISOQ)". Information assurance in computer networks: methods, models and arhitectures for Network Security, Proceedings/ International Workshop MMM ACNS 2001, St.Peterburg, Russia, May 21-23 2001, LNCS (2001), 90-101.
