Slashdot Mirror
LAMP Lights the OSS Security Way
Kevin Young wrote to mention a ZDNet article which goes into some detail on new results from a Department of Homeland security initiative. It's called the 'Open Source Hardening Project', and (funded to the tune of $1.24 Million) the goals of the initiative are to use a commercial tool for source code analysis to buck up the security base of many OSS projects. LAMP (the conglomeration of Linux, Apache, MySQL, and PHP/Perl/Python) was a 'winner' in the eyes of the project. From the article: "In the analysis, more than 17.5 million lines of code from 32 open-source projects were scanned. On average, 0.434 bugs per 1,000 lines of code were found, Coverity said. The LAMP stack, however, 'showed significantly better software quality," with an average of 0.29 defects per 1,000 lines of code, the technology company said.'"
I'm so sick of everyone making their software depend on MySQL. If you're software is any good it should be able to run on more then one DB, at least Postgres.
To me, MySQL is like the MS Access of the Open Source world.
autopr0n is like, down and stuff.
Well, once you read this snippet from the article, they'll have enough ammo:
"There is one caveat: PHP, the popular programming language, is the only component in the LAMP stack that has a higher bug density than the baseline, Coverity said."
I assume he means the baseline of 0.434 bugs/1000 lines, and that if they removed PHP from the LAMP stack, that average bug count would go down even further.
Also from the article: The lowest was the XMMS audio player, with 0.051 defects per 1,000 lines of code.
Being someone who has used Amanda for many years and also XMMS, I find it hard to believe. Amanda has few problems (unless its the tape drive itself) and XMMS crashes sometimes when you just push a button in the "wrong way".
I think there can be a big difference between actual number of bugs and the perceived number of bugs. This almost makes counts like this useless for actually comparing software.
Security is not a feature, security is design. This ultimely means that security should provide good default values, knowledge about how to prevent buffer underruns/overruns and most importantly knowledge how to use a system. This means that security only will need tools to help a system architect and developer to confront him with his limits of his human brain and have a well documented yet very simple concise system and low speed development cycles.
Open source is great because of the many eyes, knowledge sharing and having nothing to do with corporate tradeoffs (the users have the largest voice. But it stinks in the fact that any noob can make programs which are badly designed and are a serious risk to security, however someone may learn faster form the mindsharing in the open source world. To have a well concise system so much more is needed than just some bugfixes. OSS is just a proof that closed source coorporate software is not good with security, but it isn't proof of sound security.
Most interesting is OpenBSD with it's oustanding default values, it's very own high profile malloc which prevents coders for lot of buffer underrunes/overruns, outperforming other malloc implementations. It has a very high quality of manpages and if you want to do something then you have to RTFM. That's what security should be, other than some less known bugs. I would even suggest that it would be better in the name of security that people would use program derivation (which is a very concise way to do formal verification). PIE and all other solutions maybe look practical, but they don't solve the lacking attention for "secure by design".