The diversity of information sources involved in typical forensic investigations means that investigators spend 80 per cent of their time simply figuring out what information to use and how, an IBM security researcher has warned.
“A lot of time analysts spend in investigations is trying to discover the information in data sources,” Adrian Barfield, an IBM senior consultant for enterprise content management solutions, told the audience at IBM’s recent Smarter Analytics Live 2013 forum in Melbourne.
“This leaves them only 20 per cent of their time to do the real analysis work.”
That reality turns conventional models of security-related information processing on their heads – forcing organisers to sift through large volumes of information for conceptual context rather than getting stuck straight into it.
Much of that work requires bridging the analytical gap between structured and unstructured information – particularly when it comes to correlating structured activity logs with broader types of information, such as helpdesk logs or security incident reports.
“Things are becoming more and more complicated,” Barfield said. “As we create systems to try and track these things, unfortunately the criminals become more sophisticated in the ways they’re trying to break things.”
“Look at cyberterrorism and hacking,” he continued. “There are more and more break points in these systems, and we need a way to adapt as quickly as possible, to identify any patterns, put new models in place to track them, and then provide a streamlined mechanism for people to review vast amounts of information as quickly as possible.”
Organisational tools such as IBM’s UIMA (Unstructured Information Management Architecture) – an unstructured-information handling standard that IBM created for its Watson artificial-intelligence platform, then released to the open-source community as the Apache UIMA project – are emerging to provide support for forensic scientists working their way through masses of data.
“Whether we’re looking at transactions between bank accounts or at cyber-attacks, we can take it to the nanosecond or millisecond, go right down to see what is happening, and visualise that,” Barfield explained.
Because today’s unstructured data includes not only text fields but social-media content such as tweets and masses of image files, providing consistent visualisation of security and other incidents requires platforms that can deftly handle the various types of data involved.
IBM, for one, has bundled its investigation techniques into an offering called Intelligent Investigation Manager, which is designed for fraud investigation and analysis for companies in the insurance, banking and healthcare industries.
Built around the i2 Fraud Intelligence Analysis engine IBM bought two years ago, the package’s techniques highlight the increasingly important role of big-data analysis in fraud and security investigations.
With volumes of fraudulent information continuing to grow at a steady pace, Barfield said, companies need to make data reconciliation and analysis part of their everyday operations.
“In building a set of processes that I will go through to process claims and investigations, we have the ability to manage the flow of information between individuals,” he explained, “and ensure that when information flows from one point to the next, we have captured all of the necessary information before you get to that step.”
“Intelligence is really a mechanism of taking information through an analysis phase, and deriving some sort of meaning from it – and then using that outcome as the intelligence. By providing a streamlined mechanism for people to review vast amounts of information as quickly as possible, we can take these investigations through to a successful closure.”