Slashdot Mirror
Study Confirms the Government Produces the Buggiest Software
Sparrowvsrevolution writes in with a link to a Forbes story about the lackluster code produced by government agencies."Humans aren't very good at writing secure code. But they're worst at it when they're paid to do it for the U.S. government, according to a study that will be presented at the Black Hat Europe security conference in Amsterdam later this week. Chris Wysopal, chief technology officer of bug-hunting firm Veracode plans to give a talk breaking down a vulnerability analysis of 9,910 software applications over the second half of 2010 and 2011. Government-built applications came out far worse than those created by the commercial software industry or the finance industry. Only 16% of government web applications were secure by OWASP standards, compared with 24% of finance industry software and 28% of commercial software. By SANS standards, only 18% of government apps passed, compared with 28% of finance industry apps and 34% of commercial software. Wysopal and others blame the difference on a lack of accountability of federal contract developers, who aren't held to security standards and are even paid extra to fix their bugs after creating them."
Unfortunately, all the outsourcing going on in the Government (because it's easier to get money for a contract than to hire a developer on a permanent basis) is what's really killing the code here. Most outsourcing firms have a "throw the code over the wall" attitude, and spend more time deflecting blame for bugs than trying to fix them. I can't think of a business where there's less accountability than Government contracting, except possibly foreclosure management....
I was a software tester for the DoD and can confirm the stupidity here. (I can't really talk about the exact program but I can tell you with 100% certainty that it was mission critical.) We were contracted to run massive amounts of automated testing on the latest build of the software I was working on. Upon finding bugs, we needed to do regression testing... to decide if we would fix them in the latest build, because if they were present in previous versions we were under no obligation to do so unless specifically paid to do so.