How many legs does a three-legged dog have if you call a tail a leg? It still has three legs, just because you call a tail a leg does not make it a leg.
Complexity is security's biggest enemy. This adage is as old as information security itself. But stop and think about a typical enterprise: sure, they have databases, customer apps, Web apps, messaging systems, app servers, and a boatload of custom code, and all those systems create a lot of work for security teams that have to harden them for production use. However, those systems are not the most complex in the enterprise. Ask yourself - is there anything more complex in your enterprise than Kerberos or PKI or any number of security and identity protocols?
Complexity is security's biggest enemy, and yet the security mechanisms themselves are the most complex parts of the enterprise. This simple, sobering thought explains so much of security's struggle. We identify problems in some complex app, but the controls introduce yet more complexity (and a whole series of unintended consequences brought on by the interaction of two or more complex systems). Infosec is good at locating fires, problem is we're pouring gas on the fire when add yet more complexity.
The security industry has identified numerous problems in applications and servers, but the security controls brought in to remediate these issues, well, where did they come from? They did not emerge from a mountaintop in Switzerland where all the keys to perfect programming are known. Frequently they were written by the same companies and even teams who wrote the other products we're trying to protect. Often they run inside the same server code.
Even when the security products are written by a different vendor there is no particular evidence that they raise the bar in overall security. In fact they may lower it.
Veracode's State of Software Security report showed that security products have the most insecure code of all product types! Worse than learning systems, worse than customer support, worse than operations. This is a message that deserves to be told again and again. Just because you call your product a security product does not mean its adding security, it may be subtracting security.
It's even worse than that because when an isolated app fails, you have an isolated failure in which you can hopefully contain the fallout. But when a security or identity system fails, the damage is potentially widespread. Contagion, if you will. Your access management system that does authentication across 27 apps: what happens if it fails? Suddenly 27 apps have inherited a big problem.
The typical security engagement goes something like this:
As for step 5, there is no replacement for a synthetic test to identify weaknesses in the ways the different parts are integrated. This is essential. App development teams take it for granted that they have unit test code for all functionality. Security teams often throw stones at methodologies like Agile, but say this for Agile - they got the importance of testing. Do most companies have similar rigor for automated test suites on the identity and access management systems that provide access control across all users? In a word, no.
Your next Security Development Lifecycle target should not be your company's next big application project, it should be your security and identity products themselves. For those products - how much did you trust versus how much did you verify? Security, heal thyself.
Gunnar Peterson is a Managing Principal at Arctec Group Gunnar Peterson (@oneraindrop) works on AppSec - Cloud, Mobile and Identity. He maintains a blog at http://1raindrop.typepad.com. View Full Bio