#rsac #infosec #devops #sdn If you cross log analysis with infrastructure integration you get some interesting security capabilities
A lot of security-minded folks immediately pack up their bags and go home when you start talking about automating anything in the security infrastructure. Automating changes to data center firewalls, for example, seem to elicit a reaction akin not unlike that to a suggestion to putting an unpatched Windows machine directly on the public Internet.
At RSA yesterday I happened to see a variety of booths with a focus on .. .logs. That isn't surprising as log analysis is used across the data center and across domains for a variety of reasons. It's one of the ways databases are replicated, it's part of compiling access audit reports and it's absolutely one of the ways in which intrusions attempts can be detected.
And that's cool. Log analysis for intrusion detection is a good thing. But what if it could be better?
What if we started considering operationalizing the process of acting on events raised by log analysis?
One of the promises of SDN is agility through programmability. The idea is that because the data path is "programmable" it can be modified at any time by the control plane using an API. In this way, SDN-enabled architectures can respond in real time to conditions on the network impacting applications. Usually this focuses on performance but there's no reason it couldn't 'be applied to security, as well.
If you're using a log analysis tool capable of performing said analysis in near-time, and the analysis results in suspicious activity, there's no reason it couldn't inform a controller of some kind on the network, which in turn could easily decide to enable infrastructure capabilities across the network. Perhaps to start capturing the flow, or injecting a more advanced inspection service (malware detection perhaps) into the service chain for the application.
In the service provider world, it's well understood that the requirement in traditional architectures to force flows through all services is inefficient. It increases the cost of the service and requires scaling every single service along with subscriber growth. Service providers are turning to service chaining and traffic steering as a means to more efficiently use only those services that are applicable, rather than the entire chain.
While enterprise organizations for the most part aren't going to adopt service provider architectures, they can learn from then the value inherent in more dynamic network and service topologies. Does every request and response need to go through every security service? Or are some only truly needed for deep inspection?
It's about intelligence and integration. Real time analysis on what is traditionally data at rest (logs) can net actionable data if infrastructure is API-enabled. It's taking the notion of scalability domains to a more dynamic level by not only ensuring scale of services individually to reduce costs but further to improve performance and efficiency by only consuming resources when necessary, instead of all the time. The key is being able to determine when it's necessary and when it isn't.
More reading on infrastructure architecture patterns supporting scalability domains
In a service provider world that's based on subscriber and traffic type. In the enterprise it's more behavioral analysis, it's what someone is trying to do and with what application or data.
But in the end, both environments need to be dynamic with policy enforcement and service invocation based on the unique combination of devices, networks and applications and enabled by the increasing prevalence of API-enabled infrastructure.
SDN is going to propel not just operational networks as a cost savings vehicle, but as part of the technology that ultimately unlocks the software-defined data center. And that includes security.