Добавил:
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:

Information protection in digital communication systems. Textbook

.pdf
Скачиваний:
0
Добавлен:
07.09.2026
Размер:
2 Мб
Скачать
131
Each user is assigned a subset of the role hierarchy, and in each session the set of permissions of the roles that make up a fragment of this hierarchy is available. This approach makes it possible to significantly simplify access control by implicitly assigning permissions, since in real life, as a rule, users are strictly ordered by the degree of responsibility corresponding to the level of authority they have. Moreover, more trusted users who are higher on the career ladder always have all the powers of less trusted users subordinate to them. The hierarchy of roles accurately reflects this situation.
Other implementations of role-based security policies also involve introducing various restrictions on the PA, UA relationships, and user, roles, and permissions functions. The main thing for these restrictions is that they all reflect the specific distribution of powers and areas of responsibility between participants in various information processing processes. Lets look at a few examples that demonstrate the rich possibilities of using the role-based access control model:
1. Mutually exclusive roles. The set of roles is divided into subsets that combine roles that cannot be assigned to a user at the same time and are considered incompatible. Thus, a user can only be assigned one role from each subset of incompatible roles.
Mutually exclusive roles are implemented by the so-called. static separation of duties, when the conflict of incompatible powers is resolved at the stage of assigning roles. This policy is well suited for an information processing system in which users are prohibited from combining certain responsibilities. For example, in the banking system, the same user cannot simultaneously be assigned the roles of an operator responsible for performing certain operations and an auditor responsible for monitoring their implementation.
2. Restrictions on the simultaneous use of roles within one session. In this case, the set of roles is also broken down into subsets of incompatible roles, but the UA relationship can assign any combination of roles to the user. However, during a session with the system, a user can simultaneously activate no more than one role from each subset of incompatible roles.
132
Because a user can switch between different roles during a session, he must avoid incompatibility conflicts between them, this policy is called dynamic separation of duties. This policy is more flexible than static separation of concerns because it allows for more complex access control schemes to be implemented. In particular, it allows you to prohibit a user who has a significant set of roles and permissions from using them all at the same time. In certain situations, this allows you to protect yourself from a Trojan horse attack: for example, a user can be prohibited from
simultaneously accessing valuable information and running “untrusted”
programs introduced into the system by other users. Correctly selected restrictions on the simultaneous use of roles make it possible to implement control over information flows, which is generally characteristic of mandatory security models.
3. Quantitative restrictions when assigning roles and powers. This model is intended for cases where a role can only be assigned to a limited number of users and/or certain privileges can only be granted to a limited number of roles.
The meaning of these conditions is that by limiting the number of users performing certain operations, the circle of persons who are responsible for carrying out the corresponding actions is narrowed. For example, there should not be more than one administrator in the system, or the right to destroy documents can be assigned to only one role.
4. Grouping of roles and powers. Roles and powers that complement each other and that do not make sense to assign individually are combined into groups that can only be assigned as a whole. To do this, additional rules are introduced according to which any role can be assigned to a user only if a certain set of roles is already assigned to him, and a role can only be granted permission if a certain set of permissions is already associated with it.
The introduction of such restrictions simplifies system administration in cases where permissions must be granted in a specific set or when roles must be assigned in a certain sequence. For example, it makes sense to provide access to some objects (say, personal directories) only immediately
133
by both reading and writing. A typical example of role grouping is a situation where a certain user who supervises the work of other users must have powers equal to the total powers of all his subordinates, i.e. the manager role forms one group with the executor roles. It should be noted that the role hierarchy is a special case of grouping roles and powers.
Since all of the listed options for restrictions, as well as any others, can be used in various combinations, the role model is very easily adapted for each specific case, which is its main advantage over other models. Role­based policy provides wide scope for developers of access control systems: on the one hand, the use of an access rights matrix can turn it into a kind of discretionary model, but, on the other hand, the use of strict rules for the distribution of roles between sessions and users, as well as powers between roles, allows you to build on its basis a full-fledged regulatory policy. Consequently, the properties of a system built in accordance with the role model are determined solely by the nature of the restrictions used and can be in a very wide range, which does not allow a formal proof of the safety of the model for the general case.
Summarizing the properties of a role-based access control policy, it should be stated that, unlike other policies, it practically does not guarantee security using formal proof, but only determines the nature of the restrictions, the observance of which serves as a criterion for the security of the system. This approach makes it possible to obtain simple and understandable access control rules that can be easily applied in practice, but deprives the system of an evidence-based theoretical basis. In some situations, this circumstance makes it difficult to use role-based policies, but in any case, it is much more convenient to operate with roles than with subjects, since this is more consistent with common information processing technologies that provide for the division of duties and responsibilities between users. In addition, a role-based policy can be used simultaneously with other security policies, when the powers of roles assigned to users are controlled by a discretionary or mandatory policy, which allows the construction of multi-level access control schemes.
134
4.2.15. Probabilistic models
Models of this type examine the probability of overcoming a defense system within a certain time. The advantage of the models is a numerical assessment of the resistance of the protection system, the disadvantage is the initial assumption that the system can be opened.
The goal of the model is to minimize overcoming the defense system.
Game model
The model is built according to the following principle. The developer creates the initial version of the security system. After this, the attacker begins to overcome it. If by time T, at which the attacker has overcome the protection system, the developer does not have a new option, the protection system has been overcome. If not, then the process continues. That is, the model describes the process of evolution of the protection system over time [1].
Full overlap safety system model
The model pinpoints each area that requires protection, evaluates security controls, their effectiveness, and their contribution to security across the entire computing system. Each object O that requires protection is associated with a certain set of actions that an attacker can resort to to gain unauthorized access to the object. The main characteristic of a set of threats T is the probability of occurrence of each of the malicious actions. In a real system, these probabilities can be calculated with a limited degree of accuracy (Fig. 4.7).
The set of object-threat relations form a bipartite graph in which the edge <ti oj> exists if and only if ti (tiT) is a means of gaining access to object oi (oiO). The relationship between objects and threats is of a one-
to-many type, i.e. one threat can spread to any number of objects and an object can be vulnerable to more than one threat. The goal of the defense is to cover every edge of the graph and erect a barrier to access along that path.
135
Fig. 4.7. Multiple object-threat relationships
The model is completed by the third set, which includes security tools M, which are used to protect information in a computer system. Ideally, each
mk (mkM) should eliminate some edge <ti oj> from the graph in the figure. A set of Μ security features transforms a bipartite graph into a tripartite graph. In a secure system, all edges are represented as <ti mk> and <mk oj>. Any edge of the form <ti oj> defines an unprotected object. The same security feature can cover more than one threat and/or protect more than one entity. The absence of the <ti oj> edge does not guarantee complete security (although the presence of such an edge provides the potential for unauthorized access, except in the case where the probability of ti appearing is zero).
Concept of a full overlap system
A full-coverage system is a system in which there are protections for each path of entry [1].
Let:
T — set of threats;
О set of protected objects;
M security set;
V set of vulnerabilities mapping TO onto a set of ordered pairs V = {ti,O}i, representing penetration paths into the system;
136
B is a set of barriers a mapping of TOM or VM onto a set of ordered triplets {ti oj mk}, representing the points at which protection is required.
If {ti oj} V provides for {ti oj mk} B, then the jth object is protected.
The main advantage of this type of model is the ability to numerically obtain an assessment of the degree of reliability of an information security system. This method does not directly specify a model of an information security system, but can only be used in combination with other types of information security system models.
4.2.16. Information models
Stream models define constraints on the systems input/output ratio that are sufficient to implement the system. These models are the result of applying Shannons information theory to the problem of system security. These models include the non-interference and non-inferiority models.
Laissez faire model
Non-interference is a constraint in which high-level user input cannot be mixed with low-level user output. The laissez faire model views the system as consisting of four entities: high input, low input, high output, low output.
Consider a system whose output to user u is determined by the function out(u, hist.read(u)), where hist.read(u) is the input history of the system (traces) whose last input was read(u) the read command executed by the user u. Security is defined in terms of purge input histories, where purge removes commands executed by a user whose security level does not dominate u's security level.
For certain systems, the non-interference model is especially good in that if the input sequence X is not mixed with the output sequence Y and X is independent of the input of other users, then I(X,Y)=0, where I(X,Y) is the reciprocal of X and U information.
137
Irreducibility model
The non-derivability model is expressed in terms of users and information associated with one of two possible privacy levels (high and low).
A system is considered irreducibly secure if low-security users cannot obtain high-security information from any actions of high-security users. That is, information leakage cannot occur as a result of high-level users sending high-level information to low-level users.
This definition prevents low-level users from using information available to them to obtain high-level information, but does not protect high­level users from being viewed by low-level users. It simply requires that low-level users be unable to use information available to them to obtain high-level information.
4.3. INTEGRITY CONTROL MODELS
4.3.1. Biba model
Biba’s Mandatory Integrity Model
This model is often called the inversion of the Bell — LaPadula model, and therefore the basic rules of this model are simply an inversion of the
rules of the Bell LaPadula model: NRU“no read down (NRD)” and NWD“no write up (NWU)”.
An NRD rule is defined as preventing subjects from reading information from an object with a lower integrity level. An NWU rule is defined as preventing subjects from writing information to an object with a higher integrity level.
One of the advantages of this model is that it inherits many of the important characteristics of Bell — LaPadula model, including its simplicity and intuitiveness. This means that designers of real-world systems can easily
138
understand these rules and use them to make design decisions. In addition, because Biba’s mandated integrity model, like Bell LaPadula model, is based on a simple hierarchy, it is easy to explain and depict to users of the system.
On the other hand, the model is an obvious contradiction with the NRU and NWD rules. This means that if it is necessary to build a system that prevents threats to both secrecy and integrity, then the simultaneous use of the rules of the Bell LaPadula model and Biba models can lead to a situation in which the levels of security and integrity are used in opposite ways.
Subject Downgrade Model
Biba’s second model is a slightly relaxed rule of reading down.
Here the subject is allowed to read down, but as a result of such reading the level of integrity of the subject is reduced to the level of integrity of the object. This model places no restrictions on what a subject can read and assumes a monotonic progression of integrity levels.
Object Downgrade Model
Biba’s latest model is implemented by relaxing the write-up rule. The model allows writing up, but as a result, the integrity level of the object is lowered to the integrity level of the subject performing the recording. This model also places no restrictions on what a subject can read or write, it also implies a monotonic change in integrity levels, and does not contain any mechanisms for increasing the integrity level of an object.
In practical application, the Biba model relies too heavily on the concept of trusted processes. That is, the problem of the need to create trusted processes to increase or decrease the integrity of subjects or objects is very significant. It should be noted that this model does not provide mechanisms for increasing integrity, which leads to a monotonous decrease in the integrity of the system.
139
4.3.2. Clark-Wilson model
This model was created in 1987 by Clark and Wilson. Its creation was facilitated by the analysis of methods for commercial organizations to manage the integrity of their paper resources in a manual office.
Let us introduce some notation:
D — a finite data set;
CDI restricted data elements;
UDI unlimited data elements,
Moreover: D = CDI UDI, CDI UDI = .
Subjects are included in the model as a set of components that initiate transformation procedures (TPs) any non-zero sequences of elementary actions (an elementary action is a state transition that causes changes in some data elements). TPs can be represented as functions that associate a subject and a data element with a new data element as follows: TP: subjects
xDD.
TP actions that subjects (capable of changing certain data) perform on data.
This model, like other models, has its own rules. Lets look at them.
Rule 1: The system must have integrity assertion procedures (IVP (example checksum verification) asserts that a given CDI has the appropriate level of integrity) that approve any CDI.
Rule 2: Application of any PP to any CDI must preserve the integrity of the CDI.
Rule 3: Only the PP can make changes to the CDI.
Rule 4: Subjects can only initiate certain TPs over certain CDIs.
Rule 5: Appropriate policy regarding segregation of responsibilities of
entities. That is, the computer system defines a policy to prevent subjects from changing the CDI without the appropriate involvement of other subjects.
Rule 6: Some special TPs can convert UDI into CDI.
This rule allows certain TPs to receive a UDI as an input and, after a corresponding increase in integrity, issue a CDI as an output.
140
Rule 7: Each application of CDI must be registered in a special CDI, to which only information sufficient to restore a picture of the operation process of this CDI can be added. That is, the use of a special registration journal.
Rule 8: The system must recognize entities attempting to initialize the PP.
This rule defines mechanisms to prevent attacks in which one entity attempts to impersonate another.
Rule 9: The system should allow changes to authorization lists only to special subjects.
These rules define how integrity can be verified, how CDIs can be modified and by whom, and how UDIs can be turned into CDIs. That is, all changes and those who are trying to make these changes are tracked here.
The disadvantage is the difficulty of implementing VIP and methods to prevent CDI from corrupting integrity.
The main advantage of this model is that it is based on time-tested business methods for handling paper resources. Therefore, it should not be viewed as an academic study, but rather as a body of existing methods. The Clark-Wilson model also provides researchers with methods for working with integrity that differ from traditional level-based approaches such as the Bell LaPadula and Biba models.
The main disadvantage of the model is that IVP and methods for preventing CDI from corrupting integrity are not easy to implement in real computer systems.
4.4. MECHANISM FOR PROTECTION AGAINST THE THREAT OF DENIAL OF SERVICE
4.4.1. Mandate model
The mandate model includes many of the characteristics of the Bell
LaPadula and Beebe models.
Subjects of the system are assigned priorities, which may be the same,
lower or higher than the priority of any other subject. Objects correspond to