I assume offensive is focused almost exclusively within NSA and maybe a few elements of DOE/DHS (I'd be utterly terrified if every random OIG office within the government had an offensive security team, like they do have SWAT teams now...) It's also probably a case where they can have a huge lead because (for legal reasons, as well as market reasons) there's limited demand in the commercial world for "full contact" offensive security -- just pentesters and the like, or actual criminal activity, or for the lulz.
The other thing is resources other than ideas. Even I could come up with ideas like "oh, so we need to intercept all the traffic? Let's put fiber splitters at the MAEs and IXes and such.", but having the resources to subsidize facilities, work with carriers, etc. is entirely different.
This is an interesting comment because I would have thought that defense and offense would be fairly symmetrical. Is the asymmetry due to lack of interest in defense? Is it because it's easier to break things then to keep them from breaking? E.g. it's much easier to teach someone to shoot well than it is to teach someone to heal a bullet wound.
It's different when you're designing/implementing security for a new system which is very limited in scope (e.g. I would have loved to have been on the PAL design team for nuclear weapons) and very important/high profile. If you're doing anything else, it's much more a conventional IT job.
In the commercial world, it's kind of like how bitcoin (being disruptive, even if relatively undeveloped and toylike) is super popular, and is exciting, while no one is really super-enthusiastic about implementing x9.9898908098, ACH, etc.
I actually had no idea about there being different contractors on the offense side, since my contact has been really with the defense/customer side (which is shared with most of DOD). It's somewhat reassuring, as I wouldn't actually trust the "defense" people I've met with the level of information and access NSA obviously has.
I agree with you that the USG is contractorizing infosec --- though note that this makes it even easier for them to recruit --- and that contractorization degrades their capabilities.
I disagree with you on offensive security. I also think you should be aware that the contractor resources the offensive side has don't entirely overlap the defensive stuff; there are better contractors available for the offensive stuff.
And all this goes out the window when it comes to pure signals intelligence and cryptography, where we're not even close.