The IRS Mistakenly Exposed Thousands of Social Security Numbers

Another day, another slipup by the Internal Revenue Service. 

The incident involves the unwitting exposure of "tens of thousands" of Social Security numbers, according to a recent audit by the independent transparency and public-domain group Public.Resource.org. The identifying numbers were on the Internet for less than 24 hours after being discovered, but the damage was done. And unfortunately, the data-breach concerns some of the most sensitive types of transactions: Those made by nonprofit political groups known as 527s.

Every so often, 527s have to file tax forms to the IRS, which then get added to a database. The database itself is hardly a secret; the IRS has been sending updated records routinely to Public.Resource.org and other public-interest groups, and it's a favorite among political reporters. But when the IRS told the group's founder, Carl Malamud, to disregard the Form 990-Ts included in the agency's January release, he took a closer look at the files in question.

After analyzing the breach, Malamud wrote a letter to the IRS pointing out 10 instances where a social security number was accidentally revealed on the government's website—just a small sample of the larger breach. 

Just the day before, Malamud had filed another letter to the agency describing a problem with the 990-Ts. Of over 3,000 tax returns contained in the January update, 319 contained sensitive data the agency should have scrubbed, Malamud wrote in the July 1 report that he filed to the inspector general's office. In that mixup, some 2,319 social security numbers—perhaps more—were revealed.

To determine the extent of the exposure, we’ve analyzed our logs and have also analyzed the data received from the IRS. We maintain a privacy registry based on any clicks made on the privacy cover sheet on the top of each return. That registry indicates that 8 clicks were made from 4 unique IP addresses. However, none of those resulted in privacy complaints and could have been made by an automated process.

In addition, we examined our FTP and HTTP logs. We only maintain a 7-day window for HTTP logs and did not see any HTTP-based access that was not from a search engine crawler. For the FTP logs (which indicates bulk download activity), we did not see extensive activity for the January directory, but it was clear that at least one copy of the DVD ISO image (the image of the original DVD) had been transferred.

Public.Resource.org took down its copy of the compromised 990-Ts and replaced them with a clean version that the IRS had sent. But it was another day before "senior White House officials" the IRS removed the files from public view on their end, on July 3.

Calling the IRS's efforts at data security "unprofessional and amateur," Public.Resource.org isrequesting that the IRS shut down the entire 527 database to prevent further lapses. In an email, Malamud told me that the IRS has, in fact, shut down the database—but that it should also reopen it as soon as possible in the interest of transparency.

In May, the IRS drew fire for singling out conservative political groups for greater scrutiny, leading to the resignation of the agency's acting director and sparking a slew of congressional hearings.

I've called the IRS for comment, and I'll update if I hear anything.

Update: An earlier version of this post didn't make sufficiently clear the distinction between the 990-Ts and the 527 database, which are each the source of separate, if similar, problems. Both the tax documents and the database revealed social security numbers; the IRS sent Public.Resource.org a clean copy of the first but didn't fix the second until Malamud contacted the agency.