Stakeholder and Concerns Modeling the View Concepts Generic View Services Allocation
This view should be developed for security engineers of the system. It focuses on how the system is implemented from the perspective of security, and how security affects the system properties. It examines the system to establish what information is stored and processed, how valuable it is, what threats exist, and how they can be addressed.
Major concerns for this view are understanding how to assure that the system is available to only those that have permission, and how to protect the system from unauthorized tampering.
The general architecture of a 'security system' can be modeled with ADML entities (components, ports, connectors, and roles).
The subject of these are components that are secured, or components that provide security services. Additionally Access Control Lists and security schema definitions are used to model and implement security.
This section presents basic concepts required for an understanding of information system security.
The essence of security is the controlled use of information. The purpose of this section is to provide a brief overview of how security protection is implemented in the components of an information system. Doctrinal or procedural mechanisms, such as physical and personnel security procedures and policy, are not discussed here in any depth.
Figure 1 depicts an abstract view of an information system architecture, which emphasizes the fact that an information system from the security perspective is either part of a local subscriber environment (LSE) or a communications network (CN). An LSE may be either fixed or mobile. The LSEs by definition are under the control of the using organization. In an open system distributed computing implementation, secure and non-secure LSEs will almost certainly be required to interoperate.
Figure 1 Abstract Security Architecture View
The concept of an information domain provides the basis for discussing security protection requirements. An information domain is defined as a set of users, their information objects, and a security policy. An information domain security policy is the statement of the criteria for membership in the information domain and the required protection of the information objects. Breaking an organization's information down into domains is the first step in reducing the task of security policy development to a manageable size.
The business of most organizations requires that their members operate in more than one information domain. The diversity of business activities and the variation in perception of threats to the security of information will result in the existence of different information domains within one organization security policy. A specific activity may use several information domains, each with its own distinct information domain security policy.
Information domains are not necessarily bounded by information systems or even networks of systems. The security mechanisms implemented in information system components may be evaluated for their ability to meet the information domain security policies.
Information domains can be viewed as being strictly isolated from one another. Information objects should be transferred between two information domains only in accordance with established rules, conditions, and procedures expressed in the security policy of each information domain.
The concept of "absolute protection" is used to achieve the same level of protection in all information systems supporting a particular information domain. It draws attention to the problems created by interconnecting LSEs that provide different strengths of security protection. This interconnection is likely because open systems may consist of an unknown number of heterogeneous LSEs. Analysis of minimum security requirements will ensure that the concept of absolute protection will be achieved for each information domain across LSEs.
Figure 2 shows a generic architectural view which can be used to discuss the allocation of security services and the implementation of security mechanisms. This view identifies the architectural components within a LSE. The LSEs are connected by CNs. The LSEs include end systems, relay systems, and local communications systems (LCSs), described below.
Figure 2: Generic Security Architecture View
The end system and the relay system are viewed as requiring the same types of security protection. For this reason, a discussion of security protection in an end system generally also applies to a relay system. The security protections in an end system could occur in both the hardware and software.
Security protection of an information system is provided by mechanisms implemented in the hardware and software of the system and by the use of doctrinal mechanisms. The mechanisms implemented in the system hardware and software are concentrated in the end system or relay system. This focus for security protection is based on the open system, distributed computing approach for information systems. This implies use of commercial common carriers and private common-user communications systems as the CN provider between LSEs. Thus, for operation of end systems in a distributed environment, a greater degree of security protection can be assured from implementation of mechanisms in the end system or relay system.
However, communications networks (CNs) should satisfy the availability element of security in order to provide appropriate security protection for the information system. This means that CNs must provide an agreed level of responsiveness, continuity of service, and resistance to accidental and intentional threats to the communications service availability.
Implementing the necessary security protection in the end system occurs in three system service areas of TOGAF. They are operating system services, network services, and system management services.
Most of the implementation of security protection is expected to occur in software. The hardware is expected to protect the integrity of the end system software. Hardware security mechanisms include protection against tampering, undesired emanations, and cryptography.
A "security context" is defined as a controlled process space subject to an information domain security policy. The security context is therefore analogous to a common operating system notion of user process space. Isolation of security contexts is required. Security contexts are required for all applications (e.g., end user and security management applications). The focus is on strict isolation of information domains, management of end system resources, and controlled sharing and transfer of information among information domains. Where possible, security-critical functions should be isolated into relatively small modules that are related in well-defined ways.
The operating system will isolate multiple security contexts from each other using hardware protection features (e.g., processor state register, memory mapping registers) to create separate address spaces for each of them. Untrusted software will use end system resources only by invoking security-critical functions through the separation kernel. Most of the security-critical functions are the low-level functions of traditional operating systems.
Two basic classes of communications are envisioned for which distributed security contexts may need to be established. These are interactive and staged (store and forward) communications.
The concept of a "security association" forms an interactive distributed security context. A security association is defined as all the communication and security mechanisms and functions that extend the protections required by an information domain security policy within an end system to information in transfer between multiple end systems. The security association is an extension or expansion of an OSI application layer association. An application layer association is composed of appropriate application layer functions and protocols plus all of the underlying communications functions and protocols at other layers of the OSI model. Multiple security protocols may be included in a single security association to provide for a combination of security services.
For staged delivery communications (e.g., electronic mail), use will be made of an encapsulation technique (termed "wrapping process") to convey the necessary security attributes with the data being transferred as part of the network services. The wrapped security attributes are intended to permit the receiving end system to establish the necessary security context for processing the transferred data. If the wrapping process cannot provide all the necessary security protection, interactive security contexts between end systems will have to be used to ensure the secure staged transfer of information.
Security management is a particular instance of the general information system management functions discussed in earlier chapters of TOGAF. Information system security management services are concerned with the installation, maintenance, and enforcement of information domain and information system security policy rules in the information system intended to provide these security services. In particular, the security management function controls information needed by operating system services within the end system security architecture. In addition to these core services, security management requires event handling, auditing, and recovery. Standardization of security management functions, data structures, and protocols will enable interoperation of security management application processes (SMAPs) across many platforms in support of distributed security management.
Copyright © The Open Group, 1998, 2000