AOL Takes Down Site With Users' Search Data

Network News

X Profile
View More Activity
By Ellen Nakashima
Washington Post Staff Writer
Tuesday, August 8, 2006

AOL issued an apology yesterday for posting on a public Web site 20 million keyword searches conducted by hundreds of thousands of its subscribers from March to May. But the company's admission that it made a mistake did little to quell a barrage of criticism from bloggers and privacy advocates who questioned the company's security practices and said the data breach raised the risk of identity theft.

"This was a screw-up and we're angry and upset about it," the company said in a statement. "Although there was no personally-identifiable data linked to these accounts, we're absolutely not defending this. It was a mistake, and we apologize."

The posted data were similar to what the U.S. Justice Department had been seeking when it subpoenaed Internet companies, including AOL, last year. AOL complied and handed over search terms that were not linked to individuals. Google Inc. fought the subpoena in court and won.

The AOL data was posted at the end of last month on a special AOL Web site designed by the company so researchers could learn more about how people look for information on the Internet. The company removed the data over the weekend when bloggers discovered it.

The Washington Post did not review the full 439-megabyte data set but contacted bloggers who had looked at it.

For the posted data, each person using AOL's search engine was assigned a unique number to maintain anonymity, the company said. But some privacy experts said scrutinizing a user's searches could reveal information to help deduce the person's identity.

Michael Arrington, editor of the blog TechCrunch, said some of the data contained credit card numbers, Social Security numbers, addresses and names.

"People put anything they can think of into the search boxes," he said.

Based on his analysis so far, out of 20 million queries, the number that contained sensitive personal financial information such as credit card and Social Security numbers is probably "in the hundreds," he said.

"Most people aren't stupid enough to type their Social Security numbers in a search engine, but it's definitely enough to make AOL look stupid," he said.

Some bloggers said some of the information available included queries on how to kill one's spouse and child pornography.

Experts said people search for all sorts of personal data -- including their own names -- with the assumption that it will remain private.

"I search on myself," said David H. Holtzman, president of GlobalPOV, a blog and consulting firm on privacy and security and author of the forthcoming book "Privacy Lost." "Now you think you have a disease or you have some emotional issue -- I'm a single parent and I'm always looking for things. All of a sudden there's a correlation between my name and something very private that I don't expect to have dumped all over the Internet."

Kevin Bankston, an attorney with the San Francisco-based Electronic Frontier Foundation, said AOL's apology was appreciated but the damage had already been done.

"The horse is out of the barn," he said. "The data's out there and been copied. This incident highlights the dangers of these companies storing so much intimate data about their users."

The mishap was rooted in an effort by AOL to design a Web site aimed at helping researchers do their jobs more effectively by including AOL open-source data tools, company spokesman Andrew Weinstein said.

A technician posted the data to the site without running them past an in-house privacy department, not realizing the implications, Weinstein said. An internal investigation is underway to determine what happened and how to prevent future occurrences, he said.

However, Weinstein also noted that identifying an individual by search terms alone is difficult because someone could have typed in a friend's name or address instead of his own. The AOL search network had 42.7 million unique visitors in May, so the total data set covered 1.5 percent of search users that month. The 20 million search records represent about one-third of 1 percent of the total searches conducted on the AOL network in that period, the company said.

The data were gleaned from searches conducted by people with AOL user accounts in the United States.


© 2006 The Washington Post Company

Network News

X My Profile
View More Activity