Address Space Layout Randomization is an exploit mitigation technique that focuses on preventing Return Oriented Programming attacks. It’s become one of the “must have” tools for a secure program (like DEP) and it’s preset in all modern user-oriented operating systems.
Mitigating ROP is pretty important as most modern exploits take advantage of it. And ASLR would be entirely effective in an ideal world where every single part of address space is randomized and 64bit address space is impossible to bruteforce and heapspray doesn’t exist. We don’t live in that world and there are universal ASLR bypasses for Windows and Linux, heapspray does exist, and the majority of users are stuck in a 32bit address space (and 64bit vanilla ASLR isn’t necessarily impossible to bruteforce).
Windows is actually pretty on top of things with ASLR (as of Windows 8) and /FORCEASLR but there’s always going to be a way around it (unless some things seriously change.)
So what’s the answer?
Well, for non-performance critical applications perhaps a solution like Gadgetless Binaries would be a viable option. Gadgetless binaries would compile code in such a way that an attacker would be unable to make use of static address space instructions to form their attack.
There is a performance hit here so I’m not saying to compile everything with it, but for security critical applications why not? There are specific areas of Windows address space that are loaded in the same exact place every time – why not compile that area with gadgetless binaries and avoid situations like this?
Perhaps I’m just missing something. Maybe this would require paying out or some such thing but it seems like a great idea to me as ASLR isn’t going to solve every problem. At least not with current implementations (outside of PaX ASLR, which makes use of many other features via Grsecurity to prevent attacks against ASLR).