Securing The Software-Defined Network

With the emergence of software-defined networks (SDNs), security must become a core component of the network. Network security can no longer be an afterthought, or added with the assumption that it will "just work" on top of an existing network. With SDN, security services are pre-planned and become the foundation for connectivity. The benefits are centralized policy management, automated provisioning, and real-time mitigation. Sounds good, right?

The problem is the security perks you get from SDN are only as good as what you build into the system from the start. If any element or interaction in the SDN model can be compromised, the integrity of the whole network can be affected.

Integrating Security And SDNRegardless of the architectural model or the controller-agent communication protocol you choose, SDNs provide perimeter security to an organization. Yes, networks still have a perimeter in terms of today's threat landscape. The perimeter, however, is no longer a single boundary or device dividing outside from inside.

The whole concept of the Internet of everything, where any IP-based device connected to a network may pose a threat, requires each network element or function within an organization be secured in its own right. SDNs address this issue, because agents and controllers can be provisioned to provide perimeters to individual devices or services.

Security services integrated into agent systems such as identity management, threat defense, and content inspection, as well as those that impose compliance and regulatory requirements, can be applied to connections and flows. Attacks and anomalies are reported to controllers, which then enforce network-wide containment and distribute protection updates.

[Want to see the major SDN models from top vendors, in living color? Check out our slideshow on "10 Software-Defined Networking Architectures."]

Once deployed, security policy provisioning, problem resolution, and performance monitoring becomes centralized and automated.

Security Requirements For SDN

SDN can provision and enforce a security strategy, but the success of this strategy depends on how well the SDN itself is protected. A SDN consists of a number of layers and interactions, each requiring its own protection.

Deploying SDN also introduces new potential threats and vulnerabilities, as well as putting a different spin on some old favorites. The best example is a simple denial of service (DoS) attack. In a traditional network, an attacker typically targets one or more devices by overwhelming the CPU. This limits control plane services on each victim device, and can send traffic down a black hole if alternate forwarding paths are not available.

To execute a very successful DoS attack on an SDN, an attacker can simply target one agent device in the network by injecting false flows. This bogus information is then easily distributed to multiple SDN agents. It can overwhelm the controller CPU and quickly lead to the propagation of invalid flow information on all agents serviced by the controller.

The following table summarizes security requirements for the key components of an SDN:

SDNs provide an opportunity to make security the focal point of a network solution, rather than an add-on. Many existing, proven security mechanisms can be implemented and more easily provisioned and monitored using SDN. As new threats are detected or regulatory policies are required, enforcement can be simplified and streamlined. To be most successful, however, the network architects must understand the components of SDN and ensure each has its own protection.