The best definition of a security benchmarking effort I am aware of is in Chapter 11 of my book, The Pragmatic CSO, which provides a good perspective on why benchmarking is important.

Since it is very hard to have objective, defendable measures of security effectiveness, impact, etc., a technique that can yield very interesting insight into the performance of your security program is to compare it to others. If you can get a sample set of standard questions, then you can get a feel for whether you are off the reservation on some activities and out ahead of others.

Benchmarking has been in use in other IT disciplines for decades. Whether it was data center performance or network utilization, companies have always felt compelled to compare themselves to others. It’s part of the competitive, win at all costs mentality that pervades business.

So one of the best ways to figure out how good your security is, and get a feel for various other operational aspects of your security program, is to figure out how you compare to someone else. The objective is not to come up with a “security number” or “risk score”, but to present information in the context of other companies that face the same kinds of attacks. This provides management with what they always want: a perspective on the level of risk they are willing to take.

If you are behind a reasonable peer group, they can decide to invest more or to accept the risks of a less effective security program. If they are ahead, maybe they will opt to maintain or even accelerate investment in the unlikely event they can differentiate on security). Or, yes, they might decide to scale back on security ‘overhead’. Either way, it’s a win for you as the practitioner, because you know where you stand and the decision makers are actually making informed decisions with data. How novel!

But before we can start thinking about comparing all the metrics we’ve decided are important and are now collecting systematically, we need some kind of infrastructure and mechanism to share this data, safely and securely.

A few years ago I did a lot of research into building a security benchmark, and customers clearly agreed that any sharing mechanism would need to ensure:

  1. Anonymity: First and foremost, these customers wanted to make sure the data wasn’t attributed back to them. No way, no how. Of all the things I discussed with these customers, this was non-negotiable. There could be no way for another customer could identify source data or derive which company provided any of the data.
  2. Integrity: The next issue was making sure the data was meaningful. That means it must be objectively and consistently gathered. Obviously there would need to be some level of agreement on what to count and how to count it, and that would likely be the purview of a third party.
  3. Security: This goes hand in hand with anonymity, but it’s different in that potential customers need to understand how the data would be protected (at a granular level) before they’d be comfortable sharing.

Given all that, is it any wonder that security benchmarking remains in its infancy? When talking to any potential community aggregator or commercial benchmark offering, be sure to dig very deeply into how the data is both secured and aggregated to calculate the benchmarks. You need to ensure proper data encryption and segregation to make sure your data doesn’t get mixed with others, and that even if it somehow does, it wouldn’t be accessible. Additionally, you’ll want to make sure any device uploading data (this must be systematic and automated, remember) is mutually authenticated and authorized so no one can game the benchmark.

From an infrastructure protection standpoint, make sure all the proper controls are in place. Things like strong identity management, egress filtering, HIPS (if not whitelisting on the devices with access to the data), as well as significant monitoring on the network and database. Given some recent high-profile breaches, it’s not unreasonable to expect network full packet capture as well. Ultimately you need to be comfortable with how your data is protected, so ask as many questions as you need to.

From an application standpoint it’s also reasonable to expect the code to be built using some kind of secure development methodology. So learn about the threat models the vendor (or community) used to design the protection, as well as to what degree automated and non-automated testing mechanisms were used to scrutinize the application at all points during the development process. Learn about audits and pen tests, and basically crawl into very dark places in the provider’s infrastructure to get comfortable.

This is a tall order and adds substantially to the due diligence required to get comfortable participating in a security benchmark. We understand this will be too high a hurdle for some. But keep your eyes on the prize: making security decisions based only actual data, within the context of your peer group. As opposed to doing what your gut tells you, or politics, or prayer.

Once you clear this intellectual hurdle it’s time to define your peer groups for comparison and how to analyze the data. That’s next.