Hacker News new | past | comments | ask | show | jobs | submit login

See the posting above about the Arstechnica article.

During the last days of 2023 there was a big discussion, also on HN, after it was revealed that all recent Apple devices had a hardware backdoor that allowed bypassing all memory access protections claimed to exist by Apple.

It is likely that the backdoor consisted in some cache memory test registers used during production, but it is absolutely incomprehensible how it has been possible for many years that those test registers were not disabled at the end of the manufacturing process but they remained accessible for the attackers who knew Apple's secrets. For instance any iPhone could be completely controlled remotely after sending to it an invisible iMessage message.




Can't Apple just push an software update with some:

    if (user_id == "adrian_b")
       pwn ();

?


"Convenient software/hardware bugs"... but "they are not backdoors, I swear!"


> It is likely that the backdoor consisted in some cache memory test registers used during production, but it is absolutely incomprehensible how it has been possible for many years that those test registers were not disabled at the end of the manufacturing process but they remained accessible for the attackers who knew Apple's secrets.

I think we are nearly certain that the bug is because of a MMIO accessible register that allows you to write into the CPU's cache (its nearly certain this is related to the GPU's coherent L2 cache).

But I don't think it's 'incomprehensible' that such a bug could exist unintentionally. Modern computers and even more so high end mobile devices are a huge basket of complexity that has so many interactions and coprocessors all over the place I think it's very likely that a similar bug exists undiscovered unmitigated.

> For instance any iPhone could be completely controlled remotely after sending to it an invisible iMessage message.

I don't think the iMessage was invisible I think it deleted itself once the exploit had run, its also worth noting just how complicated the attack chain was and that the attacker _needed_ a hardware bug just to patch the kernel whilst having kernel code execution.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: