How Big Data is changing the security analytics landscape


 

big-data-and-analytics-the-year-ahead.jpg

Image credit: Picasa

 Storied, beautiful Edinburgh, Scotland hosted the EMEA Congress of the Cloud Security Alliance (CSA) in September 2013. CSA’s Big Data Working Group released its 2013 Big Data for Security Intelligence report at this gathering. The new
research offering outlines “how the landscape of security analytics is
changing with the introduction” of Big Data tools, as well as the
differences from traditional security analytics.

“The goal of Big Data analytics for security is to
obtain actionable intelligence in real time,” said
Alvaro Cardenas, lead author of the report in the CSA press release. “Although Big Data analytics holds significant promise,
there are a number of challenges that must be overcome to realize its true
potential. We have only just begun, but are anxious to move forward in helping
the industry understand its potential with new research directions in Big Data
security.”

An interview with two CSA members about the Big Data report

TechRepublic recently spoke with Alvaro Cardenas and Wilco van Ginkel, co-chair
of the Working Group, about the report, how
Big Data is
changing the security landscape, and what IT professionals need to know to stay
abreast of the new approaches.

TechRepublic: What are the major goals of CSA’s Big Data
Analytics subgroup?

Alvaro Cardenas: One of the goals we wanted to achieve is to
understand how Big Data is different from traditional data in the tools it provides
to us. Big Data is now a hot topic from a business perspective, bit it is hard
for a consumer to identify what is specifically unique about Big Data. One of
our main goals then was to differentiate Big Data, and then to exemplify how Big
Data is helping security in ways that other technologies were not able to do
previously, things we were not able to solve before.

Traditional vs. Big Data analytics

TechRepublic: What are the differences between traditional
and Big Data analytics?

Alvaro Cardenas: The differences are being driven by
technology, such as the
Hadoop
framework and the ecosystem around it for batch processing, stream processing,
processing data in motion for stream computing. These frameworks and the
commoditization of the data warehouses can actually now produce a big cluster
of computers, managed efficiently and cheaply. This is the way we can now
approach this problem. Before only credit card companies or telephone companies
were able to invest enough to have these Big Data warehouses and collect and
analyze historical long-term trends and correlations. But nowadays these
technologies are pretty much available to everyone interested in deploying
them. So the big changes are actually being driven by the technology and also access
to both software and hardware to manage these large-scale information
processing tasks.

Section 3.0 of the report proposes the following evolution
for data security analytics:

  • 1st
    Generation: Intrusion Detection Systems
  • 2nd
    Generation: Security Information and Event Management (SIEM). Also called
    “1st Generation SIEM”
  • 3rd
    Generation: Big Data Analytics in Security. Also called “2nd
    Generation SIEM”

The progression of data security analytics

TechRepublic: The third section of your report provides a
progression of data security analytics from the legacy perimeter approach to Big
Data capabilities. How would you describe these developments?

Alvaro Cardenas: The first generation (intrusion detection
systems) came to a close when people realized that fully protecting a system
wasn’t possible. There’s no way to perfectly protect a system or protect it
from attacks. People have been working on intrusion detection systems for
almost three decades. The first ones were very specific, very targeted — sort
of like extensions of your firewall in a way. You would create signatures, and
you would look specifically for something malicious that could be happening.
They were very specific signatures, and you would keep track of infections and
detect intrusions. One problem we realized (with first generation intrusion
detection) was we were generating a lot of false alarms.

So there’s a variety of information that needs to be aggregated
and correlated, and that’s what the second generation (SIEM) does, dealing with
the correlation of data and the false alarms. Eventually what they were doing
was allowing a centralized security operations center where people would see in
a dashboard all of the security indicators of your network. It would allow
users to collect analytics and data on trends so for example all the data
analytics that were created with this information.

We are at the birth of this third generation, what some
people are calling “second generation” SIEM. The problem with first generation
SIEM is that it does not scale very well. Sometimes you have to delete data or
keep data in different schema to allow them into the databases. So they don’t
scale very well, and they don’t allow you to incorporate several streams of
data, things that they didn’t think they had to use when they designed the
system. We are now, for example, monitoring websites and text-based tweets or
the message of emails. Unstructured data is very difficult to capture with
first-gen SIEM technologies. One of the advantages of Big Data and NoSQL
databases is that they can store these data in a format that is scalable and at
the same time they allow you to create queries to understand the data better.
So we found out that this is the promise of Big Data, moving security
information and monitoring to the next level.

Big Data security analytics: what you need to know

TechRepublic: Pretend I am a CISO at a Fortune 1000 company.
What are the most important things I need to know about Big Data security
analytics in order to stay on top of the technology?

Alvaro Cardenas: Big Data enables various capabilities, for
instance, forensics and the analysis of long-term historical trends. By
collecting data on a large scale and analyzing historical trends, you would be
able identify when an attack started, and what were the steps that the attacker
took to get ahold of your systems. Even if you did not detect the original
attack in your systems, you can go back and do an historical correlation in your
database and systems to identify the attack. So long-term historical analysis
is one advantage.

Another is the efficiency of queries. So when you want to
understand your data, Big Data allows you to carry out complex queries and
receive results in a timely fashion.

Finally, I think it’s the difference in technology between
batch processing and streaming data. You probably need both in your systems.
Streaming data is just analyzing data online without doing these historical
correlations, just streaming the traffic. So this would be a tool to identify
more pressing attacks that appear suddenly, whereas batch processing is better
for analyzing long-term trends. So those are some of the concepts that I think
are important.

Wilco van Ginkel: If I can add a little something. This is
indeed a great discussion to have, but there’s one question that has to be
asked first, that’s what I would do if a CISO approached me. That is: What is
the concern of the CISO at the moment, and how can Big Data analytics help him
or her? It’s not so much do we need to jump on the Big Data analytics
bandwagon, yes or no? The question is: Why should I? So is there anything in
the risk profile or the security status of the company that bothers him or her?
If so, let’s talk about Big Data analytics as a potential solution. If yes,
then we have the discussion. There’s a lot of talk about Big Data, but it’s in
such an embryonic stage that a lot of companies, just like with cloud five
years ago, jump on it without understanding why they do it in the first place, and
what’s the benefit. That would be my first discussion with a CISO.

A lot of companies are just scratching the
surface of Big Data analytics. They’re doing small proofs of concept. They’re
trying to jump from, say, descriptive analytics, which is all about what’s
happened in the past, to prescriptive analytics, what’s going to come. But it’s
all very small scale — they’re just trying to get their head around it and the
data. So it’s really in its infancy stages, there’s no change in the industry. It’s
coming from analytical startups especially focused on security analytics. From
the startup market, you see them trying to tackle security analytics from a
different angle. In the enterprise market, a lot of companies are thinking
about it, but don’t have a real idea about it, with a few exceptions. Again,
they’re just trying to figure out if it’s really something for them and what is
the value-add.

Thank you to Alvaro Cardenas and Wilco van Ginkel for making time for this interview.

 

This entry was posted in Big Data. Bookmark the permalink.