Cisco Security Group Access: An Introduction

The concept of identity-based network access -- essentially, providing credentials to gain access to a network or a resource -- is not a new phenomenon. Over the years, it's evolved from logging onto individual resources, to single sign-on (SSO), and to centralized user repositories that make use of multiple authentication methods and enforce authorization policies customized by user, group, access type and access method.

Fundamentally, IBNA is granting a level of trust to a user or device based on proof of identity. Security within an organization is based on control, enforcement and predictability. The more that is known about the users on a network -- individuals or devices -- the more comprehensive and effective the security policy will be. Implementing IBNA is a great way for organizations to review their network design, secure it and optimize it because it poses questions such as “Who are you?” and “Where do you need to go and why?”

Understanding network flows and user-resource requirements is the basis for Cisco’s Security Group Access (SGA), which falls under the umbrella of Cisco TrustSec. SGA offers several benefits:

1)Grants network resource access based on identity and associated policy.

2)Enhances security and control as traffic flows are more easily segmented.

3)Reduces the cost and complexity associated with large firewall policies and access-list rules.

4)Provides a mechanism for consistent and dynamic policy propagation across different platforms.

5)Enables centralized policy management and auditing per identity.

SGA is based on the concept of assigning a user a Security Group Tag (SGT). The SGT is a 16-bit value inserted into an 802.1AE frame. Assignment of the SGT is known as classification. When a user is authenticated using 802.1X, MAC Authentication Bypass (MAB) or Web Authentication (WebAuth), the SGT is assigned as part of the authorization policy on the AAA server, usually the Cisco Identity Services Engine (ISE).

Additionally, network resources such as infrastructure devices (routers, switches, firewalls, wireless LAN controllers (WLCs), and services are also assigned SGTs. Tags may also be manually defined at the port level or statically mapped to IP addresses, subnets or VLANs on individual network devices.

SGA relies on the propagation of SGT information to devices that become policy enforcement points. Propagation of the SGT may be done inline by special hardware ASICs that tag each traffic flow sourced from a user with the assigned SGT. This embedded value is carried with the flow and examined by an enforcement point.

In the case of devices not capable of handling the embedded SGT, tag information can be propagated using SGT eXchange Protocol (SXP), which is a TCP-based peer-to-peer connection where the “speaker” propagates IP address to SGT binding information to a “listener.” The listener may use this information to enforce policy based on learned SGT values.

[Humans are the weak link in security, but blaming them for clicking on links or opening phishing emails overlooks basic human nature. Read Michele Chubirka's analysis of the problem in "The Real Reason IT Security Fails."]

As the user traffic passes through an enforcement point, the assigned SGT is must be permitted by a Security Group ACL that has been downloaded from the centralized policy store (ISE), or statically configured on the enforcement device.

SGACLs simplify policy and reduce rule complexity as illustrated by the simple example below.

In general, because of the syntax used to create rules based on IP address:

(# of sources) * (# of destinations) * permissions = # of ACEs (access control entries).

So 4 subnets requiring access to three Web servers using HTTP and HTTPS would require:

4 * 3 * 2 = 24 ACEs.

Using SGTs, we can assign all “users” in each subnet to one group (source), and combine the three Web servers into a secure group (destination), then apply the two permissions (permit HTTP, permit HTTPS) plus the catch-all action for non-compliant traffic:

1 * 1 * 3 = 3 ACEs.

Obviously, the more granular the SGT assignment, the more rules, but the concept is that using SGT’s provides a more efficient and meaningful way of grouping resources based on role and function.

Uniform policy enforcement is achieved by propagating SGT information and having devices apply policy consistently across the network topology. The more centralized the policy management, and the more devices that can dynamically download applicable rule sets, the less chance for configuration errors resulting from managing large numbers of statically defined ACEs across multiple network devices.

Migration to a SGA solution can be done gradually. Start with classification and propagation of SGTs between Cisco devices. SXP, which uses TCP as a transport, will enable tag propagation across non-SGA capable devices. This will facilitate the introduction of enforcement on strategic network devices such as access switches and WLC’s at the network edge as well as protection of critical resources, such as Nexus switches in the core.