DARPA wants machines to collate streams of information to free humans for judgment calls.
The Defense Department hopes to offload some of the work of analyzing network vulnerabilities to a machine, Pentagon officials said on Friday.
The Cyber Targeted Attack Analyzer is intended to reduce the workload for the department’s short-handed cyber forces by organizing information from “disparate network data sources” to more easily see computer abnormalities, according to the Pentagon’s laboratory. Information technology development efforts will kick off with a briefing for prospective contractors on Jan. 30, Defense Advanced Research Projects Agency officials said. DARPA expects to release a solicitation for project proposals within a few weeks afterward.
The trick will be reeling in all that intelligence from devices that are not necessarily compatible.
“Changing the way the information in the IT infrastructure is acquired, processed and made available” will help tackle the “scale-of-data” problem, DARPA officials said in announcing the industry event.
The ideal technology would need to automatically index data sources without much human intervention, according to officials. The CAT program “seeks to reduce the amount of time cyber defenders in the Department of Defense spend discovering cyber-attacks by federating and correlating” dissimilar data streams, officials said.
But the tool also must allow humans to exercise judgment so that they can, for instance, “query relationships between any connected data fields across the network” to probe irregularities further, officials added.
The contract labor likely will be divided into two sections, with one effort focused on research to devise a functioning system, and the other concentrating on testing to ensure the envisioned technology works and is secure.
The CAT program is one of several ongoing big data projects at DARPA related to cybersecurity.
The Pentagon is plowing $250 million annually into initiatives aimed at harnessing large data sets at the agency, the National Security Agency’s code-cracking division, and elsewhere. For example, DARPA’s Anomaly Detection at Multiple Scales, or ADAMS, system looks for willful or inadvertent malicious actions taken by trusted individuals -- insider threats -- against a backdrop of normal network activity. Another experimental technology, the Video and Image Retrieval and Analysis Tool, or VIRAT, scouts for dangerous combinations of videotaped activities captured by battlefield sensors that Defense would never have the time or people to review.
CORRECTION: The original version of this story had the wrong dollar figure for the amount the Pentagon is devoting to harnessing large data sets. The correct figure is $250 million.