Model Driven Security Policy Automation

On this blog, ObjectSecurity co-founder and CEO Ulrich Lang discusses security policy automation and model-driven security. The aim of this blog is to advocate advance the state of the art in this area through exchange of ideas. - -

Friday, 1 April 2011

Implementing security policy automation: Free lunch? Unfortunately not.

Good security is ultimately about figuring out what should happen, and making sure that anything else does not happen. In security terms, this means figuring out enterprise security and compliance policies first, then figuring out how to implement controls across technology, processes, and people.
Unfortunately this is hard, which is why most security products and methods in the market avoid enforcing your policy altogether. For example, antivirus, anti-malware, etc. are useful "hygiene" tools but do not know enough about your business to even enforce the policies that matter (e.g. PCI, HIPAA, NERC/FERC, Common Criteria...). Other tools (IDSs, compliance monitoring etc.) also do not know the policy that matters and simply monitor something, so some administrator - if they can weed through the overload - may spot that you got hacked, which is better than nothing but does not prevent being hacked. Other tools enforce a policy (e.g. firewalls, identity management), but usually not the policies or the granularity/contextuality that matters to the business. While I am a proponent of "defense in depth", I would sum the current state of most of the security vendor landscape and end-user purchasing behavior as "solving something we can solve", rather than actually solving the real security problems.
However, doing this right by stating what you want and enforcing it is hard: for example, manually producing many complex, context-aware technical policy rules ("whitelisting")  for a highly interconnected, large Service Oriented Architecture (SOA) or cloud mash-up is highly error-prone and expensive, and is also totally unmaintainable. There is also little assurance that the configured policy actually capture the intent.
Policy automation tools such as OpenPMF make this easier and more maintainable, especially for agile IT landscapes (incl. SOA/cloud) - it lets security and compliance specialists capture policies at an intuitive level as models (similar to enterprise architecture and business process models), and automatically takes care of generating/enforcing/monitoring the matching technical rules. However, this is no free lunch either - figuring out and capturing the requirements and configuring everything is not easy and takes time. It also will not work elegantly for each and every kind of system. However, when you compare it to the two alternatives (1) solving something but not the problem and (2) incurring a manual administration nightmare, it is a compelling approach.

No comments: