• Aucun résultat trouvé

Access Control Models

Dans le document A valuable extension to the (Page 105-108)

There are three access control models primarily in use today. These models are commonly used to control users rights to various resources. These resources can range from high-level business processes to low-high-level system object access. These access control models also range in administrative overhead from practically nothing to highly resource intensive (at one or more stages), depending on the amount of initial or recurring oversight they require. The corresponding level of security they provide also ranges from practically nothing to complying with best security practices. The three models are as follows:

Discretionary Access Control (DAC) Allowing access controls to be

confi gured by the respective data owners. Data owners have complete control over the fi les they create. This control model is the default for Linux.

Mandatory Access Control (MAC) Denying data owners’ full control of the data they create and having access controls managed and confi gured by administrative personnel. Users and fi les are assigned a security level and users can only access fi les having a security level that is equal to or less than their own.

Role-Based Access Control (RBAC) Assigning access to data, applications, and business processes based on users’ roles within the organization or their individual function over the data, applications, and business processes.

When performing an audit and making decisions about implementing the best access control model(s) for various resources, sufficient thought must be given to the value of the resources being secured and the cost required to secure them. This thought process should, however, take into consideration not only the value of the resources being secured, but also the impact on the organization and its customers if compromised.

The value of the item being protected is not necessarily limited to its inherent intrinsic value. Appropriate weight must be given to its perceived value and cost to the company in goodwill and/or reputation if it were compromised, as well as impact on its customers.

Chapter 4: Local Access Control

71

For instance, even if a company resource does not contain highly sensitive data (from a regulatory perspective, such as credit card or bank account numbers), but is publicly accessible or may contain data embarrassing to its customers, it is prudent to harden it from intrusion. Company message boards are a good example. They seldom contain truly sensitive data, but if a message board server is compromised, the only part of the headline “Company X Message Board Hacked” that people will remember is “Company X Hacked.”

Furthermore, customers often post sensitive items to technical message boards, revealing vulnerabilities, particularly if the boards are security- or information-technology-infrastructure related. If such a message board were compromised, vulnerabilities regarding customers could be identified and possibly exploited.

Discretionary Access Control DAC is the simplest access control model, has the lowest administrative overhead, and provides the lowest level of security. It is based on the assumption that owners should be allowed to control their own data. Owners have a free and unfettered ability to provide any level of access to others (or not) and are also free to directly (or accidentally) create, modify, and delete any of their own data.

They are also free to modify and delete the data of other owners to which they have been assigned sufficient permissions. The only real safeguards against data loss are user responsibility, a good backup scheme, and/or data recovery (or computer forensic) software. This access control model is probably the most common due to its ease of use and lack of administration, thus contributing to the success of the data recovery software industry and also providing easy targets for hackers.

The only security features or access controls implemented in this model are those configured by the data owners. By placing full control with data owners, there is an implied trust that the data owners will make wise and prudent decisions in their stewardship over the data, applications, or business processes.

For these reasons, this access control model should not be used for business-critical or sensitive data processes. Time has proven that data owners are not necessarily the best stewards over sensitive and business-critical data. In environments where DAC is used for sensitive resources, data loss and unauthorized (or unintended) access are common.

Mandatory Access Control MAC is a newer and much more sophisticated access control model than DAC, requiring more significant administrative configuration and control. It takes full control of data away from owners and places it with administrative personnel who assign owners only the level of access required.

It prevents owners from granting a less restrictive permissions assignment to resources than was assigned by administrative personnel and prevents users at one level from accessing data at a different level. Furthermore, data owners are no longer free to grant permissions to others; what they are able to do with their own data is often restricted. For instance, many implementations of MAC enable owners to create data, but not delete it, protecting owners from themselves as well as others.

Essentially, MAC protects data, processes, and applications from misuse, abuse, simple mistakes, or malfeasance. Its function lies somewhere between a patch for

ignorance and a shield against attack. Its goal is to create a carefully planned architecture where the required level of access is specifically granted for each resource, thereby creating an environment where users have the permissions they need—nothing more, nothing less.

This model requires a tremendous amount of administrative overhead to make the necessary configurations to various resources, whether they are file systems, devices, applications, or business processes. Administrative personnel need to thoughtfully and methodically map how users will access resources and grant appropriate access accordingly.

In newer distributions of Linux, a MAC feature is built into the kernel. Various Linux distributions have other, more specific MAC packages that provide enhancements over the features added to the kernel

Role-Based Access Control RBAC is a newer, alternative access control model leveraging the strength of granular access configuration created by MAC, but providing greater scalability through the creation of roles. Instead of assigning users specific access to resources, resource assignments are made to the various types of roles that exist within an organization. Users are assigned a specific role for each resource access requirement.

Roles should be set up specifically for each system, application, or device. They should not span multiple systems, applications, or devices. It is bad form and not considered best practice to assign a single role to multiple systems, applications, or devices. A substantial number of permission permutations already exist within a single system and trying to combine the permissions of several heterogeneous systems within a single role becomes cumbersome and is rife with problems.

To increase the granularity of role assignments and provide greater configurability:

• Users can be assigned numerous roles.

• Roles can be assigned multiple users.

• Roles can be assigned numerous permissions.

• Permissions can be assigned to multiple roles.

This effectively creates a many-to-many relationship granting all possible permission permutations, but through a simplified methodology of grouping the permissions assignments into corresponding roles.

This approach requires significantly more time to set up in order to completely map out the various roles that apply to resources, as well as all of the resources that exist within an environment. But once the resources have been defined and roles created, the time required to grant new users specific access rights across various resources in the environment is substantially reduced.

Various secure Linux distributions and patches that provide a variety of RBAC measures are available. As of the 2.6 kernel, Security Enhances Linux, by the NSA, or SELinux, has been built into the Linux kernel and provides measures for RBAC.

Chapter 4: Local Access Control

73

GRSecurity also has several patches available for download that minimize configuration for creating a robust RBAC system.

Chrooting

The amount of work that goes into securing a system can be partially mitigated by taking advantage of the chrooting abilities built into certain applications or by using the chroot feature that is included or can be compiled into Linux. Chroot is a combination of two words: change and root. As the name implies, chrooting changes the root directory of logged-on users or applications. It creates a sandboxed, virtual directory that is used to provide a user or an application access to only a limited subset of resources.

Certain daemons, such as FTP and SSH, have the built-in or add-in ability to sandbox users in a carefully crafted “chrooted” environment. This provides users with an emulated and simplified file structure and includes only the executables, libraries, configuration files, and so on, as needed.

More specifically, when users log in to a chrooted system, they are not actually allowed to peruse the computer’s real file system. The root directory they are able to view is really a subdirectory that has been assigned to them and includes all of the executables and dependencies needed to perform their intended functions. Theoretically, the chrooted users cannot gain direct access to areas outside the chrooted, or sandboxed, environment. However, some hard links between files or directories within a chrooted environment to files or directories outside the chrooted environment may exist, leading to users’ ability to escape the chroot jail in ways that would not otherwise have been available to them, if symbolic links had been used instead.

Chrooting for applications like OpenSSH, however, is quite a bit easier than other applications, as OpenSSH initializes itself first and performs chroot() later. This means that a less comprehensive chroot environment is necessary. Other applications are chrooted in a variety of other ways, mainly through the use of a configuration option or with the chroot() command-line tool.

Similar to FTP’s and SSH’s native (or add-in), configurable chrooting ability, many daemons provide a similar ability, except it is intended only for the user account that the daemon runs as. This means that if attackers gain control of a chrooted daemon, they are limited to the sandboxed environment. Apache is a good example of a daemon that has this built-in ability and is commonly used to protect web servers.

Numerous other server applications, particularly network listening web applications, have the ability to run in a chrooted environment. However, for any of these applications to function properly, all of their configuration files and dependencies must be copied into the chrooted environment in the same directory structure as would exist on the normal file system.

Dans le document A valuable extension to the (Page 105-108)