Does there ever come a point at which the accelerating combinatorial complexity of this context transcends human aptitude to comprehend and mitigate the hyper-inflating entropy of the problem-space? I imagine that as long as human brains, incentives and motivations are involved, we might at least be able to infer and loosely forecast the kinds of TTP (Tactics, Techniques and Procedures) in use.
If (and when, if not already occurring) autonomous machine intelligence really starts to gain leverage in offensive cyber actions, this whole game levels up, yet again. Due to the implicit advantage attackers have in regard to retrospective or system-introspective protective practices, I wouldn’t be surprised to (eventually) observe all defence deployed as autonomous offence. In future, there may be no other way to play this game.
An endemic orientation towards indefinite-extensibility and recursive logical insight suggests that the effervescing complexity only ever provides advantage for itself and as such, all retrospective analyses – although critically important – are necessarily rendered impotent by the speed and sophistication of technical metamorphosis. Once having stepped onto and into this system, the only way to survive is to run (at least) as fast as everyone else.