IT Security From The Eyes Of Data Scientists
As IT security leaders try to base more of their day-to-day decisions on statistical analysis of relevant data coming from IT infrastructure and business processes, they’re running into a skills and resource gap. Often times security teams have lots of specialists with deep technical knowledge of attack techniques and trends, but they frequently lack the skills to aggregate and manipulate data in order to draw meaningful conclusions from statistical trends. As the speed and volume of security data continues to mount, so will that gap, which is why many within the industry believe that in the coming years an IT security team will not be complete without at least one data scientist among its ranks.
“In the past, it’s always been us who’s been behind the game, trying to catch up with the attackers’ techniques,” says Dan Mitchell, product manager of data sciences for RSA The Security Division of EMC. “I think data science gives us the opportunity to get ahead of the attackers and have them be behind for a change.”
Mitchell is among a growing legion of data scientists growing active within the IT security community and one of several that Dark Reading caught up with to get their views on the value that their colleagues bring to the table, why enterprises need to employ more and how organizations can develop talent and embed these experts into their security practices.
The complex chain of techniques that attackers today use to infiltrate IT resources and steal data makes it absolutely critical that security teams spot trends and connect behaviors that span across IT infrastructure, user groups and geographical locations.
In order to do that, it requires security to have experts that can manipulate data, visualize it and draw conclusions from it. Not only that, the team needs to be able to build infrastructure to store data, normalize it and develop modeling that can answer the burning questions security analysts have about anomalies that may indicate compromise—and that infrastructure should preferably be designed to do it all automatically.
This is the exact kind of expertise a data scientist brings to the table, says Ram Keralapura, data scientist for Netskope, a cloud apps analytics and policy creation company., who explains that the CISO and data scientist have the opportunity to form a symbiotic relationship.
“Security officers have a very good understanding of the outcome they want and have identified their problems—they want to know specific kinds of information about certain kinds of anomalies or activities that are happening in their enterprise, but they don’t always know how to get that information,” says Keralapura. “Data scientists are the right people to bridge this gap and provide the insights that these security officers need in order to make more informed decisions.”
What’s more, Mitchell explains that someone with his type of expertise can help break down a lot of the silos that currently exist in the security realm.
“So, because the security industry has become so fractionalized in terms of specialty areas, data science offers a way to bring specific domain expertise and then combine that with things like machine learning, mathematical modeling and manipulating data to solve problems that extend across all specialties,” he says. “It’s really about creating the whole picture.”
[How do you know if you’ve been breached? See Top 15 Indicators of Compromise.]
Whereas in the past a lot of the mathematical minds in security tended to gravitate towards specialties like encryption or authentication, Mitchell believes that many will be diverted into data science.
“There’s so much more we can do mathematically to solve our problems,” he says. “I think you’re going to see more and more of that. It’s a larger trend.”
Many vendors have already been leading the trend of hiring and training more data scientists to develop analytics-based security products, but the role of the data scientist should also be a staple within enterprise IT security teams.
“The reason I think that businesses also have to be hiring data scientists is that in security especially, a large component of the practice is data about your particular environment,” says Michael Roytman, data scientist for Risk I/O, a vulnerability threat monitoring vendor. “A lot can be done to use that data to narrow down where you should be focusing on your security risks and that’s where an in-house data scientist plays a part.”
And, says Keralapura, it really should be a full-time role. There are several big reasons for this, he says. First, in order to develop predictive models about the enterprise’s specific data, the data scientists need to develop long-term relationships with security experts on staff and deal with data on a day-to-day basis. Second, in order to accomplish real-time detection, they’ll need to be around to help with response in real-time. And third, a full-time data scientist is crucial to helping forensics problems that could pop up at any time.
“When a problem happens and you need to look at data right away in order to identify what it was, why did it happen, how did it happen, and all of these different dimensions that need to be answered,” says Keralapura. “These things keep happening all the time.”
As enterprises seek out those with a data science background, there are two big skill sets they should be looking for. The most obvious is a high degree of mathematics and statistical analysis. The second is the coding chops of a hacker.
“You are going to want people that have some hacking ability to put things together quickly. A lot of it is going to be about changing the view quickly and some developers may know how to program well in a long development cycle,” says George Ng, data scientist for YarcData, a Cray company that focuses on graph analytics. “But if someone is trying to steal your data, the pattern isn’t something you already have in production to look for—it’s something you develop on the fly.”
Next page: The insider data scientist