The problem of ensuring the information security of an automated process control system (APCS) is considered. An overview of the main regulatory documents on ensuring the safety of automated process control systems is given. For the operative solution of the tasks of ensuring information security of the automated control system of technological processes it is proposed to use an intelligent decision support system (DSS). An example of the construction and implementation of decision rules in the composition of the DSS based on the use of neurofuzzy models is considered.
communications, energy, banking and other areas of the financial market, fuel and energy complex, in the field of atomic energy, defense, rocket and space, mining, metallurgical and chemical industries, or organizations that provide the interaction of these systems are subject to mandatory protection, in order to ensure their sustainable functioning when conducting against their computer attacks.
According to the law should be a categorization of CII objects, compiled a national register of significant CII objects, provides for the implementation of mandatory requirements to ensure the safety of significant CII objects, controlled by the state.
In order to concretize the requirements provided by the Federal law 187-FZ, and the conditions for
their use, the FSTEC of Russia issued a sub-legal regulatory base:
• The Order of Russia FSTEC No. 235 dated by 21.12.2017 “On approval of requirements for
the creation of security systems for significant objects of the critical information infrastructure
of the Russian Federation and for ensuring their functioning” [
However, these changes in the legal base, currently there are no formal methods and techniques of qualitative and quantitative assessment of the level of protection and choice of effective countermeasures to ensure full compliance with regulatory requirements to ensure comprehensive information security APCS. This doesn't allow one to fully counteract the influence of a wide range of possible cyber threats on the information resources of organizations and enterprises.
Therefore, it is urgent to develop decision support algorithms, the use of which would improve the efficiency of information security APCS of a particular enterprise.
The purpose of the study in this article is to develop the structure of the decision support system (DSS), implementing the risk assessment of information security APCS.
To achieve this goal should be solved by the following tasks: 1. Development of the DSS structure applicable to the construction of a secure APCS; 2. Development of decision support algorithm in the task of assessing the requirements to ensure the protection of information in the APCS based on artificial intelligence technologies; 3. Evaluation of the possibility of using DSS by an example illustrating the features of the application of the proposed algorithm to the construction of the APCS.
The solution of the tasks of ensuring information security of the APCS has its own characteristics. This is primarily: • high uncertainty of initial information and complexity of its receipt; • the need to consider many of the requirements for information security when evaluating and choosing the best options.
The article [
When forming the requirements for information security in the APCS by the Order of the Russia FSTEC, it is implied that it is necessary to develop a model of information security threats. It should contain a description of the APCS and current threats to information security.
As a result of analyzing the processes of ensuring information security of an APCS, the following main tasks can be identified, solved with the help of DSS: • accumulation and systematization of information on information security of APCS; • assistance in developing recommendations for minimizing the possible information security risks of APCS.
Under the risk refers to the expected potential damage from the impact of an actual threat through vulnerability to a specific information (material) asset.
At the first stage of risk analysis, the level of security of the APCS [
Fuzzy neural networks ANFIS are used for quantitative risk assessment.
The process of analyzing the security of automated systems is different in that when assessing information risks, fuzzy values in the form of expert assessments are often used as input data. This necessitates the use of fuzzy models, the main advantage of which is associated with the possibility of using for their development much smaller amounts of information about the simulated system, compared with traditional mathematical models. In this case, the information may be approximate, fuzzy in nature.
To implement a comprehensive risk assessment, it is necessary to build a system of fuzzy production rules. At the same time, the three-factor risk formula is used as a base: Risk (damage) = Threat * Vulnerability * Information (material) resource, where Threat * Vulnerability = Degree of realization of the threat (threat).
The indicators used in the development of DSS for ensuring information security of the APCS can be
both quantitative and qualitative. Therefore, there is always uncertainty in making decisions on
assessing the risks of information security of the APCS. In this case, to determine the level of
information security risk of the APCS, it is proposed to use data mining technologies using a modular
(ensemble) neural network [
The general architecture of the DSS to ensure information security of the APCS is shown in Figure 1.
The module of preliminary data processing on the APCS leads the input values of the modular neural network to a single scale. The inputs of a fuzzy neural network are given indicators of identified information security vulnerabilities of the APCS (T1 ÷ T4), indicators of the value of information contained in the system (CI), as well as the degree of possible damage in case of breach of confidentiality (C), integrity (I) or availability (A). The neural network determines the probability of realization of the threat P, after which, based on a set of rules, it assesses the risk Y and determines the security class (K) of the APCS. The output data of the rule relevance assessment subsystem are 1∗ – vector of assessments of the contribution of rules to the formation of an assessment of the probability of threats, 2∗ – vector of assessments of the contribution of rules to the formation of an information security risk assessment.
It is assumed that all rules work to some extent, i.e. have a different level of activity. However, exceeding a certain threshold value indicates a significant contribution of certain rules to the result. The selected rules can show which of the parcels are the most suitable and therefore lead to the result.
Based on the obtained security class and risk assessment Y, taking into account the contribution of decision rules to the definition of this assessment, recommendations are made in determining the composition of information protection measures.
The adoption of the correct and timely decision to ensure information security of APCS directly depends on the completeness and correctness of the established rules base. It contains solutions to one or another problem on information security of the APCS, based on the analysis of the subject area and the knowledge of experts. Therefore, the creation of a rules base in the design of DSS is a primary task.
Decision rules can be represented in a fuzzy rules base in the Mamdani fuzzy inference system and have the following form:
: If 1 is 1 and 2 is 2 and … and is , then is , where – j-th rule (j = 1,2,…,m); – input variable, (i = 1,2,…,n); – the result of applying the j-th rule and – terms (fuzzy subsets).
An important task of the study is to map the set of decision-making tasks to ensure the information security of an APCS on a set of decision-making rules.
The impact of vulnerability on the implementation of a specific threat is reflected in the rules that have the following scheme:
IF Vulnerability – HIGH, THEN the probability of threats – HIGH, etc.
According to this principle, the number of rules of the module for assessing the probability of threats implementation will depend on the number of vulnerabilities, differentiated according to the degree of danger and determining the impact of this threat.
To determine the security class, three input indicators of confidentiality (C), integrity (I) and availability (A) were introduced; at the output of the neural network, we obtain the security class of the APCS.
Input indicators of confidentiality (C), integrity (I) and availability (A) are determined by three linguistic terms, which are evaluated by an expert on a scale from 0 to 1:
L – (0; 0,3) – “Low damage”; M – [0,3; 0,7] – “Middle damage”; H – (0,7; 1] – “High damage”.
The security class of the APCS (К), which depends on confidentiality, integrity and availability indicators, is also determined by three linguistic terms, the values of which are determined using a neural network based on the rules established by an expert:
L – [0; 0,3] – “First security class”; M – (0,3; 0,7) – “Second security class”; H – [0,7; 1] – “Third security class”.
Rules that determine the security class of the APCS listed in Table 1. Since at the input we have three variables I, A and C, defined by three linguistic terms L, M and H, the rule table contains 33 =27 rules.
Similar to the definition of the class of security APCS, Figures 2 and 3 show the membership functions and rules that are used to determine the likelihood of realization of threats of information security of APCS.
In the course of the work, the compiled system of rules was implemented in the FuzzyToboxbox package of mathematical modeling in Matlab based on the ensemble of neuro-fuzzy networks ANFIS.
The structure of the ensemble of neuro-fuzzy networks, proposed for solving the problem of information security risk assessment of APCS, is presented in Table 4 and Figure 4:
The training and test samples are based on the basis of expert assessments and contain options for solving the set tasks based on the analysis of the subject area. For the construction of training, sets can also be used data from systems included in the system of information security APCS.
The detailed structure of the network that determines the probability of threats is shown in Figure 5.
The first layer is the terms of the input variables T1÷T4. Input data in this layer is converted to fuzzy. At the output of the layer, we obtain the degree of belonging to the input variable value to a certain term.
The second layer is the antecedents of fuzzy rules. Each node of this layer corresponds to one fuzzy rule. In this case, to determine the probability of threats to be realized, the inference system has 81 rules. The output node of the layer is the degree of execution of the rule μi(T).
The third layer is the conclusion of the rules. The nodes calculate the contribution of the corresponding rule to the network output.
The fourth layer is the combination of the result obtained according to different rules. The node of this layer summarizes the contributions of all the rules.
One of the drawbacks of fuzzy models is the effect of "retraining". The model gives a minimal error on the elements of the training set with a large error on the elements of the testing set. To overcome this drawback, the initial sample is divided into two subsets: training and test.
Increasing the dimension of the input vector of linguistic variables exponentially increases the
number of elements of layer 2 of the ANFIS model, as well as the number of adjustable weight
coefficients. This makes it difficult to train the model and increases the requirements for the training
sample [
This paper uses the generation of a fuzzy model using fuzzy clustering (FCM) to reduce the
dimension of the second layer of the neuro-fuzzy model ANFIS. When using FCM clustering, it is
possible to control the sample size (the number of observations in the sample, the number of input
variables) used to build the model, and the type of model (Sugano / Mamdani) [
The implementation of the rule system in the Fuzzy Toolbox mathematical modeling package is shown in Figure 6.
In order to formulate recommendations for reducing the level of information security risk in an APCS, all the rules in this case work to varying degrees. But exceeding the threshold value allows you to select the rules with the most significant contribution to the result. An analysis of the parts of the antecedents of the ranked list of rules makes it possible to identify linguistic variables and their meanings that lead to the current result.
As a result of the research, a DSS was built, which allows to assess the level of information security risk of the APCS and issue recommendations for its minimization.
The example of using the developed DSS is illustrated by the example of the APCS, the physical architecture of which is shown in Figure 7. DSS helps to assess the practical implementation of the requirements of information security standards and to ensure the necessary level of security for APCS.
The input data for the DSS are the results of the information security audit conducted at the
security facility, which includes security analysis to search for vulnerabilities, analysis of the
documentation, structure and configuration of the system. The initial data for the audit are including
the results of using network security scanners and the accumulated data of intrusion detection
systems [
Suppose that according to the results of the audit, the APCS has the following vulnerabilities: T1 – Lack of identification and authentication of subjects and objects of access; T2 – Incorrect Default Permissions; T3 – No perimeter protection of APCS, connection with corporate networks and the Internet; T4 – Lack of protection against denial of service attacks. The input parameters of the neural network are defined as follows: T1 = 0.3 (M); T2 = 0.1 (L); T3 = 0.6 (M); T4 = 0.95 (H).
At the same time, the value of information processed and circulated in data transmission networks of the APCS is defined as CI = 0.6.
At the output of the neural network, we obtain the values of the probability of the threat realization (P), the information security risk assessment of the APCS (Y), and the vector of assessments of the contribution of rules to the formation of the probability of the threat ( 1∗) and the risk estimate ( 2∗).
As the calculations showed, the probability value of the threat realization is 0.727. This suggests that the probability of a threat acting through these vulnerabilities is above average. In turn, the inputs of a fuzzy neural network to determine the level of information security risk of the APCS are given the values P = 0.727 and CI = 0.6. With such input indicators at the output of the network, we obtain the value of the risk level equal to 0.537, that is, the risk level is also above average.
Formation of recommendations for reducing the level of information security risk of the APCS is as follows. All the rules in this case work in various degrees but exceeding the threshold value (in this case it is equal to 0.95) allows you to select the rules with the most significant contribution to the final result. These selected rules allow us to show why such an assessment of probability and risk was obtained, and to identify weak points. Based on knowledge of weak points, appropriate measures are taken to protect information from the APCS.
The DSS ultimately issues an information security risk assessment for the APCS, and recommendations for ensuring a given level of information security with instructions on what should be paid priority attention.
The proposed structure of the DSS to ensure information security of automated process control systems.
The algorithm for decision support based on data mining technology using a modular (ensemble) neural network has been developed, which allows solving the problem of risk assessment and compliance of requirements for ensuring information security of an APCS and identifying current threats to a specific protection object.
Risk assessment of information security of APCS is necessary to develop recommendations for reducing the risk level and choice of effective countermeasures that ensure full compliance with the regulatory requirements for ensuring comprehensive information security of the APCS.
The use of the proposed fuzzy neural network in assessment information risks of the APCS makes it possible to adequately use the qualitative and quantitative evaluations obtained from the experts as input data and will also improve the level of enterprise security by maintaining the information protection system of the APCS up to date.
This work is partially supported by the Russian Science Foundation under grants №17-48-020095.