SEAR and the Apple team does an excellent job of security on iOS, and should be commended greatly on that.
Not only are they willing to develop hardware features and plumb that throughout the entire stack, they're willing to look at ITW exploits and work on ways to mitigate that. PPL was super interesting, they decided it wasn't 100% effective so they ditched it and came up with other thigs.
Apple's vertical makes it 'easy' to do this compared to Android where they have to convince the CPU guys at QC or Mediatek to build a feature, convince the linux kernel to take it, get it in AOSP, get it in upstream LLVM, etc etc.
Pointer authentication codes (PAC) is a good example, Apple said f-it we'll do it ourselves. They maintained a downstream fork of LLVM, and built full support, leveraged in the wild bypasses and fixed those up.
One of the knock on benefits of this too is increased security across all platforms as long as someone exercises that code path on one of apples new processors with a hardened runtime.
In theory it makes it easier to catch stuff that you can’t simply catch with static analysis and it gives you some level of insight beyond simply crashing.
I buy Apple products not just because they do a great job with security and privacy, but because they do this without needing to do it. They could make plenty of money without going so deep into these features. Maybe eventually it’d catch up with them but it’s not like they even have competition forcing them to care about your privacy.
Their commitment to privacy goes beyond marketing. They actually mean it. They staffed their security team with top hackers from the Jailbreak community… they innovated with Private Relay, private mailboxes, trusted compute, multi-party inference…
I’ve got plenty of problems with Apple hypocrisy, like their embrace of VPNs (except for traffic to Apple Servers) or privacy-preserving defaults (except for Wi-Fi calling or “journaling suggestions”). You could argue their commitment to privacy includes a qualifier like “you’re protected from everyone except for Apple and select telecom partners by default.”
But that’s still leagues ahead of Google whose mantra is more like “you’re protected from everyone except Google and anyone who buys an ad from Google.”
If you have it enabled, then every thirty seconds (regardless of whether you’re actively on a call), your phone will make a request to a signaling server owned by your mobile ISP. So if you’re on T-Mobile and traveling in some other country with no cell service, but you’re connected to WiFi, then T-Mobile will see your public IP address. (IIRC, this also bypasses any VPN Profile you have enabled on your device, because the signaling system is based on a derivative of IPSec that could have problems communicating over an active VPN tunnel.)
I found out about this when I was wiresharking all outbound traffic from my router and saw my phone making these weird requests.
Apple actually does warn you about this in the fine print (“About WiFi calling and privacy…”) next to the toggle in Settings. But I didn’t realize just how intrusive it was.
I know my mobile ISP can triangulate my location already, but I don’t want to offer them even more data about every public IP of every WiFi network I connect to, even if I’m not roaming at the time.
And after all that hardcore engineering work is done, iMessage still has code paths leading to dubious code running in the kernel, enabling 0-click exploits to still be a thing.
None of that ran in the kernel. Everything happens within a single process up until the sandbox escape, which isn't even covered in your article. The article's sequel* goes into detail about that part, which involves subverting a more privileged process by exploiting logic errors to get it to execute code. The only involvement by the kernel is passing IPC messages back and forth.
Google could have added MTE for a couple of years now, but apparently don't want to force it on OEMs as part of their Android certification program, it is the same history as with OS updates.
to be fair, most of MTE's benefit is realized by having enough users running your apps with MRE enabled, rather than having it everywhere.
This is because MTE facilitate finding memory bugs and fixing them - but also consumes (physical!) space and power. If enough folks run it with, say Chrome, you get to find and fix most of its memory bugs and it benefits everyone else (minus the drawbacks, since everyone else has MTE off or not present).
trade offs, basically. At least on pixel you can decide on your own
They do that now because they care about your security, but to make it difficult to modify (jailbreak) your own devices to run your own software that is not approved by Apple.
What they do is against your interests, for them to keep the monopoly on the App Store.
Whenever I read about it, I am surprised at the complexity of iOS security. At the hardware level, kernel level, all the various types of sandboxing.
Is this duct tape over historical architectural decisions that assumed trust? Could we design something with less complexity if we designed it from scratch? Are there any operating systems that are designed this way?
>Is this duct tape over historical architectural decisions that assumed trust?
Yes, it's all making up for flaws in the original Unix security model and the hardware design that C-based system programming encourages.
> Could we design something with less complexity if we designed it from scratch? Are there any operating systems that are designed this way?
Yes, capability architecture, and yes, they exist, but only as academic/hobby exercises so far as I've seen. The big problem is that POSIX requires the Unix model, so if you want to have a fundamentally different model, you lose a lot of software immediately without a POSIX compatibility shim layer -- within which you would still have said problems. It's not that it can't be done, it's just really hard for everyone to walk away from pretty much every existing Unix program.
It’s been designed with lower user trust since day one, unlike other OSes of the era (consumer Windows, Mac’s classic OS).
Just how much you can trust the user has changed overtime. And of course the device has picked up a lot of a lot of of capabilities and new threats such as always on networking in various forms and the fun of a post Spectre world.
I think that there's also inherent trust in "hardware security" but as we all know its all just hardcoded software at the end of the day, and complexity will bring bugs more frequently.
Security in this context means the intruder is you, and Apple is securing their device so you can't run code on it, without asking Apple for permission first.
That makes no sense for a phone because you go outside with it in your pocket, leave it places, connect to a zillion kinds of networks with it, etc. It's not a PC in an airgapped room. It is very easy for the user of the device to be someone who isn't you.
Any sufficiently secure system is, by design, also secure against it's primary user. In the business world this takes the form of protecting the business from its own employees in addition to outside threats.
SEAR and the Apple team does an excellent job of security on iOS, and should be commended greatly on that.
Not only are they willing to develop hardware features and plumb that throughout the entire stack, they're willing to look at ITW exploits and work on ways to mitigate that. PPL was super interesting, they decided it wasn't 100% effective so they ditched it and came up with other thigs.
Apple's vertical makes it 'easy' to do this compared to Android where they have to convince the CPU guys at QC or Mediatek to build a feature, convince the linux kernel to take it, get it in AOSP, get it in upstream LLVM, etc etc.
Pointer authentication codes (PAC) is a good example, Apple said f-it we'll do it ourselves. They maintained a downstream fork of LLVM, and built full support, leveraged in the wild bypasses and fixed those up.
One of the knock on benefits of this too is increased security across all platforms as long as someone exercises that code path on one of apples new processors with a hardened runtime.
In theory it makes it easier to catch stuff that you can’t simply catch with static analysis and it gives you some level of insight beyond simply crashing.
I buy Apple products not just because they do a great job with security and privacy, but because they do this without needing to do it. They could make plenty of money without going so deep into these features. Maybe eventually it’d catch up with them but it’s not like they even have competition forcing them to care about your privacy.
Their commitment to privacy goes beyond marketing. They actually mean it. They staffed their security team with top hackers from the Jailbreak community… they innovated with Private Relay, private mailboxes, trusted compute, multi-party inference…
I’ve got plenty of problems with Apple hypocrisy, like their embrace of VPNs (except for traffic to Apple Servers) or privacy-preserving defaults (except for Wi-Fi calling or “journaling suggestions”). You could argue their commitment to privacy includes a qualifier like “you’re protected from everyone except for Apple and select telecom partners by default.”
But that’s still leagues ahead of Google whose mantra is more like “you’re protected from everyone except Google and anyone who buys an ad from Google.”
What is non-private about Wi-Fi calling?
If you have it enabled, then every thirty seconds (regardless of whether you’re actively on a call), your phone will make a request to a signaling server owned by your mobile ISP. So if you’re on T-Mobile and traveling in some other country with no cell service, but you’re connected to WiFi, then T-Mobile will see your public IP address. (IIRC, this also bypasses any VPN Profile you have enabled on your device, because the signaling system is based on a derivative of IPSec that could have problems communicating over an active VPN tunnel.)
I found out about this when I was wiresharking all outbound traffic from my router and saw my phone making these weird requests.
Apple actually does warn you about this in the fine print (“About WiFi calling and privacy…”) next to the toggle in Settings. But I didn’t realize just how intrusive it was.
I know my mobile ISP can triangulate my location already, but I don’t want to offer them even more data about every public IP of every WiFi network I connect to, even if I’m not roaming at the time.
And after all that hardcore engineering work is done, iMessage still has code paths leading to dubious code running in the kernel, enabling 0-click exploits to still be a thing.
That's one way to look at it, but if perfection is the only goal post then no one would ever get anywhere.
What's the dubious code?
Running something in the kernel is unavoidable if you want to actually show stuff to the user.
In ~2020, it was:
Attacker sends an imessage containing a PDF
imessage, like most modern messaging apps, displays a preview - which means running the PDF loader.
The PDF loader has support for the obsolete-but-part-of-the-pdf-standard image codec 'JBIG2'
Apple's JBIG2 codec has an exploitable bug, giving the attacker remote code execution on the device.
This exploit was purchased by NSO, who sold it to a bunch of middle eastern dictatorships who promptly used it on journalists.
https://googleprojectzero.blogspot.com/2021/12/a-deep-dive-i...
None of that ran in the kernel. Everything happens within a single process up until the sandbox escape, which isn't even covered in your article. The article's sequel* goes into detail about that part, which involves subverting a more privileged process by exploiting logic errors to get it to execute code. The only involvement by the kernel is passing IPC messages back and forth.
* https://googleprojectzero.blogspot.com/2022/03/forcedentry-s...
Disable iMessage via Apple Configurator MDM policy and enable Lockdown Mode.
I imagine the latter is sufficient.
PS: make sure you remove that pesky "USB accessories while locked allowed" profile that Configurator likes to sneak in.
Need an open-source MDM profile policy linter.
[dead]
Why would a nation-state actor need access to your kernel when all the juicy stuff[0] is in the iMessage process it's already loaded into?
[0] https://xkcd.com/1200/
Google could have added MTE for a couple of years now, but apparently don't want to force it on OEMs as part of their Android certification program, it is the same history as with OS updates.
Don't the Pixels have MTE? Definitely GrapheneOS does, at least to some extent.
Kind of, you need to enable it on developer tools, also Pixels are Google's, not the other OEMs.
https://developer.android.com/ndk/guides/arm-mte
https://source.android.com/docs/security/test/memory-safety/...
to be fair, most of MTE's benefit is realized by having enough users running your apps with MRE enabled, rather than having it everywhere.
This is because MTE facilitate finding memory bugs and fixing them - but also consumes (physical!) space and power. If enough folks run it with, say Chrome, you get to find and fix most of its memory bugs and it benefits everyone else (minus the drawbacks, since everyone else has MTE off or not present).
trade offs, basically. At least on pixel you can decide on your own
They do that now because they care about your security, but to make it difficult to modify (jailbreak) your own devices to run your own software that is not approved by Apple.
What they do is against your interests, for them to keep the monopoly on the App Store.
It can be both things, security and user lock in, those are orthogonal goals.
Loosely related, but they also announced an increase in their bug bounty program during Ivan Krstić's Keynote at Hexacon. [0]
[0] https://security.apple.com/blog/apple-security-bounty-evolve...
Whenever I read about it, I am surprised at the complexity of iOS security. At the hardware level, kernel level, all the various types of sandboxing.
Is this duct tape over historical architectural decisions that assumed trust? Could we design something with less complexity if we designed it from scratch? Are there any operating systems that are designed this way?
>Is this duct tape over historical architectural decisions that assumed trust?
Yes, it's all making up for flaws in the original Unix security model and the hardware design that C-based system programming encourages.
> Could we design something with less complexity if we designed it from scratch? Are there any operating systems that are designed this way?
Yes, capability architecture, and yes, they exist, but only as academic/hobby exercises so far as I've seen. The big problem is that POSIX requires the Unix model, so if you want to have a fundamentally different model, you lose a lot of software immediately without a POSIX compatibility shim layer -- within which you would still have said problems. It's not that it can't be done, it's just really hard for everyone to walk away from pretty much every existing Unix program.
> seL4 is a fast, secure and formally verified microkernel with fine-grained access control and support for virtual machines.
https://medium.com/@tunacici7/sel4-microkernel-architecture-...
It's missing "the rest of the owl", so to speak, so it's a bit of a stretch to call it an operating system for anything more than research.
Vulnerabilities are inevitable, especially if you want to support broad use cases on a platform. Defense-in-depth is how you respond to this.
iOS is based on MacOS is based on NeXT is a Unix.
It’s been designed with lower user trust since day one, unlike other OSes of the era (consumer Windows, Mac’s classic OS).
Just how much you can trust the user has changed overtime. And of course the device has picked up a lot of a lot of of capabilities and new threats such as always on networking in various forms and the fun of a post Spectre world.
why not do both :)
I think that there's also inherent trust in "hardware security" but as we all know its all just hardcoded software at the end of the day, and complexity will bring bugs more frequently.
[dead]
Security in this context means the intruder is you, and Apple is securing their device so you can't run code on it, without asking Apple for permission first.
That makes no sense for a phone because you go outside with it in your pocket, leave it places, connect to a zillion kinds of networks with it, etc. It's not a PC in an airgapped room. It is very easy for the user of the device to be someone who isn't you.
It can be both.
Any sufficiently secure system is, by design, also secure against it's primary user. In the business world this takes the form of protecting the business from its own employees in addition to outside threats.
Have they fixed regular, unencrypted connections on updates and apps launch yet?
https://sneak.berlin/20231005/apple-operating-system-surveil...