@bryanjclark the writeups on those exploits are fascinating. Like the one that devised an entire hardware architecture on top of a tiny image operator exploit, and they built a compiler to write code for it
@bryanjclark I believe it’s a combination of: 1. Image containers offer a lot of places to hide bits here and there. Had a buddy get into Steganography back in the early 2000s.
2. Images are one of the few files that iOS will download and parse when it’s pushed to you so they can put them in notifications and such.
@bryanjclark this one, god I am glad I was not misremembering it. As I replied to you I was thinking “did I hallucinate that? Sounds like science fiction”
It’s not:
@bryanjclark wasn't there a whole out-of-process architecture thing a few releases ago that was supposed to eradicate this vector all together?
@bryanjclark followup
“Fix OOB write in BuildHuffmanTable”… “ The patch suggests that it was possible to overflow the Huffman table when decoding an untrusted image…”