If you go down to the bottom of the article they actually provide the link to the Apple Update page[1]. The actual vulnerability report page [2].
They also have a zeroday report for LibTiff so I guess they found something similar in the library.
There's the links I was looking for. "Investigating further we see RAX is pointing to the very end of a heap block yet there is still a lot of data to be written as RDX (counter) is still 0xFE." Why can't Forbes write in such plain English?
They aren't. When you receive an MMS your iPhone runs code to save and display the content. From my understanding of the article, the code responsible for handling an incoming MMS contains a critical security bug enabling attackers to hide instructions in MMS content capable of reading any memory on the iPhone, even memory stored outside the Messages sandbox (like your passwords).
The bug is in the ImageIO library which would affect SpringBoard, Messages, Safari, and most other applications which process images. I don't understand how having code execution in either of these processes would give up that information unless you explicitly had texted that info.
I suppose the idea is that it leads to a full system compromise, which would include either direct access to items like wifi passwords or access to the system as you are inputting the passwords.
It's the fact that the code can read any memory, not just app memory, according to GP.
That's now how sandboxing works. You'd need an additional bug to escape the sandbox.