[SC-L] BSIMM: Confessions of a Software Security Alchemist (informIT)

[Gary McGraw]

Hi Steve,

Sorry for falling off the thread last night.  Waiting for the posts to clear was not a great idea.

The top N lists we observed among the 9 were BUG lists only.  So that means that in general at least half of the defects were not being identified on the "most wanted" list using that BSIMM set of activities.  You are correct to point out that the "Architecture Analysis" practice has other activities meant to ferret out those sorts of flaws.

I asked my guys to work on a flaws collection a while ago, but I have not seen anything yet.  Canuck?

There is an important difference between your CVE data which is based on externally observed bugs (imposed on vendors by security types mostly) and internal bug data determined using static analysis or internal testing.  I would be very interested to know whether Microsoft and the CVE consider the same bug #1 on internal versus external rating systems.  The difference is in the term "reported for" versus "discovered internally during SDL activity".

gem

http://www.cigital.com/~gem


On 3/18/09 6:14 PM, "Steven M. Christey" <coley at linus.mitre.org> wrote:



On Wed, 18 Mar 2009, Gary McGraw wrote:

> Many of the top N lists we encountered were developed through the
> consistent use of static analysis tools.

Interesting.  Does this mean that their top N lists are less likely to
include design flaws?  (though they would be covered under various other
BSIMM activities).

> After looking at millions of lines of code (sometimes constantly), a
> ***real*** top N list of bugs emerges for an organization.  Eradicating
> number one is an obvious priority.  Training can help.  New number
> one...lather, rinse, repeat.

I believe this is reflected in public CVE data.  Take a look at the bugs
that are being reported for, say, Microsoft or major Linux vendors or most
any product with a long history, and their current number 1's are not the
same as the number 1's of the past.

- Steve