US suspends vast ADVISE data-sifting system

From late 2004 until mid-2006, a little-known data-mining computer system developed by the US Department of Homeland Security to hunt terrorists, weapons of mass destruction, and biological weapons sifted through Americans' personal data with little regard for federal privacy laws.

Now the $42 million cutting-edge system, designed to process trillions of pieces of data, has been halted and could be canceled pending data-privacy reviews, according to a newly released report to Congress by the DHS's own internal watchdog.

Data mining to help fight the war on terror has become an accepted, even mandated, method to provide timely security information. The DHS operates at least a dozen such programs; intelligence agencies and the Department of Defense employ many others.

But ADVISE (Analy­sis, Dissemina­tion, Visu­ali­zation, Insight and Semantic Enhance­ment) was special. An electronic omnivore conceived in 2003, it was designed to ingest information from scores of databases, blogs, e-mail traffic, intelligence reports, and other sources, government documents and researchers say.

Sifting that enormous mass at lightning speed, ADVISE was to display data patterns visually as "semantic graphs" – a sort of illuminated information constellation – in which an analyst's eye could spot links between people, places, events, travel, calls, and organizations worldwide.

Report: DHS didn't follow guidelines

Yet ADVISE, whose existence and scope were first detailed by the Monitor in February 2006, seems to have run afoul of its own ambitious scope. It failed to incorporate federal privacy laws into its system design. From its earliest days, the system's pilot programs used "live data, including personally identifiable information, from multiple sources in attempts to identify potential terrorist activity," but without taking steps required by federal law and DHS's own internal guidelines to keep that data from being misused, the DHS Office of Inspector General (OIG) said in a June report to Congress, which was made public Aug. 13.

In a rebuttal attached to the report, the DHS Directorate for Science and Technology disagreed with most of the OIG's findings. "The ADVISE tool set is little more than an empty framework to which data must be applied," wrote Jay Cohen, DHS undersecretary for science and technology, in a letter accompanying the rebuttal. He said no privacy laws were violated.

Even in searching for terrorists, data-mining programs are supposed to ensure that Americans' personal information is used only when necessary and lawful – and only for specific and proper uses. One problem is that even data that look anonymous aren't necessarily so. For instance, even when names and Social Security numbers are stripped from data files, programmers can still identify 87 percent of Americans through their date of birth, gender, and five-digit Zip Code, researchers say. So a system has to be carefully designed and use encryption and other computer techniques to comply with the law.

Last week the Pentagon shut down its TALON terrorism database program, which had been found to hold files on peace activists. In 2003, another military data-mining project – the Total Information Awareness project – was also ended following a congressional uproar over privacy fears.

Congress last fall ordered its Government Accountability Office to audit the program for privacy and effectiveness. It asked the OIG to do the same. In February, the GAO recommended a full-blown data-privacy review of the ADVISE system. Without that, its report said, ADVISE holds "potential for erroneous association of individuals with crime or terrorism and the misidentification of individuals with similar names."

In his report to Congress, publicly released earlier this month, DHS Inspector General Richard Skinner revealed that:

1 | 2

Next