You hear it at every conference and in the halls of every university computer science program. It's mentioned in every sales pitch for cybersecurity tools and outsourcing services: There simply aren't enough qualified cybersecurity professionals.
The problem is that this statement assumes that the process of securing systems and enterprises is just fine as it is. That simple claim assumes that we need each and every one of the jobs posted, listed, and forecasted.
We don't have a workforce shortage problem. What we have is an automation-in-the-wrong-place problem. It's not about training people to do traditional network security. What we need are mathematical models that meaningfully predict risk and provide pathways to reduce it.
This lesson is easily seen in vulnerability management, but it's applicable to other fields. Think of it this way: The typical enterprise network has millions of vulnerabilities. On median, our research found that out of about 500 enterprises, IT teams fix 10% of those vulnerabilities, though some exceptional performers patch 25% on a monthly basis.
If companies were to hire enough people to eliminate every vulnerability from their systems, they'd need to at least quadruple their workforce devoted to the task. And that, of course, assumes that the rate at which vulnerabilities are found and disclosed stays constant, which it doesn't — it's constantly increasing. Does that seem reasonable? It is not.
But another reason lies in the nature of vulnerabilities themselves. For the vast majority of Common Vulnerabilities and Exposures (CVEs), the risk of exploitation is entirely theoretical. That is, nobody has weaponized the vulnerability with an exploit.
Traditionally, enterprises have treated vulnerability management as a manpower and triage problem. They assembled lists of CVEs that their scans turned up and argued over which ones to patch. Every day the list grew, and every quarter CISOs tried to make the case for additional hires.
Application of data science to this problem has shown that companies can — and do — make meaningful risk reductions with available resources. That's because the small cadre of hackers capable of developing new exploitations are highly likely to follow well-worn patterns. A complete analysis of decades of threat data bears this out.
Threat actors are more likely to develop exploits for certain vulnerabilities than for others. They look for CVEs that target assets in widespread use, which makes operating systems from Microsoft riskier than Apple, and they target vulnerabilities that allow for remote code execution more frequently than other Common Weakness Enumerations. Dozens of factors, all of which are publicly available, drive risk or reduce exploitation risk. Identification of these factors forms the basis for risk-based vulnerability management.
Just 5% of vulnerabilities pose a risk of exploitation, which means that even average organizations, in theory, have twice the capacity to patch vulnerabilities in a way that drastically reduces risk of intrusion. We wouldn't know that vulnerability management is a math problem, and not a workforce problem, without data science to prove it.
That risk-based, data-driven approach is quite suitable for other cybersecurity disciplines. User behavior and analytics tends to generate a significant amount of data that can be marshalled in service of identity management. That’s just one example.
The key is to find tools, datasets, and statistical methodologies that can help you separate the signal from the noise. The right tools will help you quantify risk and apply that analysis to prioritize the actions that get the most meaningful results. Which is to say, find tools that help you get the most out of your available resources.
If you can't find the tools, invent them. The cybersecurity community can't hire its way out of the manpower shortage, but there might just be a new startup idea in solving it — the good ones always come from practitioners automating themselves out of a job.
Machines are well suited to the task of defending networks. They can automate analysis in a way that fills the manpower gap. For CISOs and other executives facing a manpower shortage, it’s imperative that they accept this so they can adopt strategies that deal with the world as it is.Michael Roytman is the Chief Data Scientist at Kenna Security, and has spoken at RSA, BlackHat, SOURCE, Bsides, Metricon, Infosec Europe, and SIRAcon. His work focuses on cybersecurity data science and Bayesian algorithms, and he served on the boards of the Society of ... View Full Bio