Homeland Security officials are testing a supersnoop computer system that sifts through personal information on U.S. citizens to detect possible terrorist attacks, prompting concerns from lawmakers who have called for investigations. The system uses the same data-mining process that was developed by the Pentagon's Total Information Awareness (TIA) project that was banned by Congress in 2003 because of vast privacy violations. A Government Accountability Office (GAO) investigation of the project called ADVISE -- Analysis, Dissemination, Visualization, Insight and Semantic Enhancement -- was requested by Rep. David R. Obey, Wisconsin Democrat and chairman of the House Appropriations Committee. The investigation focuses on whether the program violates privacy laws, and the findings will be released after completion of the Iraq war supplemental spending bill, possibly as early as this week, a panel aide said. The ADVISE and TIA data-mining projects rely on personal data to track individual behavior and consumer transactions to develop computer algorithms that create a pattern that some behavioral scientists say can predict terrorist behavior. Data can include credit-card purchases, telephone or Internet details, medical records, travel and banking information. Privacy concerns prompted lawmakers on both sides of the aisle to introduce legislation in January to require that government agencies disclose data-mining practices in regular reports to Congress. "A serious discussion on the implications of data-mining programs is long overdue," Sen. Russ Feingold, Wisconsin Democrat and a sponsor of the bill, said yesterday. Sen. John E. Sununu, New Hampshire Republican, is also a bill sponsor. "Many Americans are understandably concerned about the idea of secret government programs analyzing their personal information. Congress needs to know more about the operational aspects and privacy implications of data-mining programs before these programs are allowed to go forward," Mr. Feingold said. A spokesman for the Department of Homeland Security did not return a call for comment. Congress also tucked language inside Homeland Security's spending bill in September requiring an investigation by the agency's inspector general, but allowed $40 million in funding to go forward in this year's budget. "The ADVISE program is designed to extract relationships and correlations from large amounts of data to produce actionable intelligence on terrorists," the spending bill said. "A prototype is currently available to analysts in Intelligence and Analysis using departmental and other data, including some on U.S. citizens." According to a Congressional Research Service (CRS) report in March 2003, TIA planned "to use data mining technologies to sift through personal transactions in electronic data to find patterns and associations connected to terrorist threats and activities." "Recent increased awareness about the existence of the TIA project provoked expressions of concern about the potential for the invasion of privacy of law-abiding citizens by the government, and about the direction of the project by John Poindexter, a central figure in the Iran-Contra affair," the CRS report said. "While the law enforcement and intelligence communities argue that more sophisticated information gathering techniques are essential to combat today's sophisticated terrorists, civil libertarians worry that the government's increased capability to assemble information will result in increased and unchecked government power, and the erosion of individual privacy," the report said. ADVISE was initiated in 2003 following the demise of the TIA project. The new system includes data-mining tools to digest "massive quantities of information from many different sources" to find "hidden relationships in the data," according to a 2004 report by Sandia National Laboratories and Lawrence Livermore National Laboratory on a Homeland Security workshop that outlined this and other technology under development. The technology is expected to analyze more than 3 million "relationships" or connections per hour, says the report, which included an example of how friends, family members, locations and workplaces can be linked by pinging the data.