In the world of cyberthreat defence, security experts and vendors have conceded that it is a matter of "when," not "if," an enterprise will be breached.
As a number of studies have found, antivirus solutions fail to detect as much as 60 percent of malware attacks. Beyond that, the 2011 Verizon Data Breach Report, 86 percent of breached organizations failed to detect that their networks were hacked, and had to be notified by a third party.
So, while everyone still advocates "layered" security to keep thieves and other attackers out, the current emphasis is on detecting and confronting those who make it inside.
That is the pitch from LogRhythm, the Colorado-based firm that announced this week an enhancement of its SIEM (Security Information and Event Management) 2.0 Big Data security analytics platform.
LogRhythm cofounder and CTO Chris Petersen said the new release offers the industry's first multi-dimensional behavioural analytics (MDBA) capability - a method of detecting anomalies in the normal, or "baseline," activities of an enterprise, across all its entire digital environment. The premise is that if it is not normal, it is probably malicious.
"This is a continuation of our original vision to collect all the data - log, flow, events across all hosts and devices - anything that tells us what's happening," Petersen said. We collect and centralise it for analytics, so we can ask questions of it.
"MDBA is an important and profound step, because it lets us go further into big data and ask questions across that data set," he said. "We ask what kinds of behaviors are being observed and look for correlation patterns in that data. Any anomalies might indicate risk, compliance or security problems."
Petersen said the different analytic techniques lend themselves to answering different types of questions. One of them is "behavioural whitelisting," which observes the normal activities of everything from apps to host servers, individual users and those using VPNs. "From those, we build up a whitelist," he said.
The platform also analyses trends, such as the amount of authentication activity from foreign countries, or the number of unique files an employee downloads on a normal day. "If we suddenly see 10 times the norm, it puts an alert on that," Petersen said.
This kind of "heuristic" behavioral analysis, which looks for activities outside the norm, has been in use for some time and, according to some analysts, has some downsides. Taylor Thomas, writing at TopTen Reviews, notes two of them:
"Obviously this sort of scanning and analysis can take some time, which may slow down system performance," Thomas wrote. "[But] the main concern with heuristic detection is that it often increases false positives ... when the antivirus software determines a file is malicious (and quarantines or deletes it) when in reality it is perfectly fine and/or desired."
[Bill Brenner in Salted Hash: McAfee smells a Shady RAT - A lot of 'em, actually]
Petersen said behavioural analysis "becomes meaningless with too many false alarms." A company press release notes that while first-generation SIEMs use behavioural analysis, it is one-dimensional and, in many cases, manual.
"In either scenario, IT and security personnel remain blind to much of the behaviour of today's advanced hackers because the evidence of their activities are buried amidst massive volumes of false positive security events, or they're miscategorised altogether as benign or 'normal' activities," the release said.
Petersen said MDBA addresses that by seeking to corroborate one type of suspicious behaviour with one or more others.
"We might see that a user is not only downloading large number of files, but using a new process that we never observed before. When you have highly corroborated events that all deviate from a pattern, the possibility that it's a false alarm is very low," he said.
Still, how does MDBA detect malicious activity if an attacker is already on the inside when SIEM 2.0 begins an analysis? Wouldn't that malicious activity be viewed as part of the "baseline" of normal activity?
"That is a challenge," Petersen said. "But you can address that by build off the users who you know are legitimate. We observe the behaviour of privileged users, and then apply that behaviour to all the others. We find out which ones are acting differently and why. It's the same for hosts - the web servers that are public facing. We'll use the known servers to build a model for whitelisting."
He said the reality that some breaches are inevitable requires a "shift in mindset." When an attacker gets inside, the focus has to be on "detect, contain and eradicate," he said.