How would an exploit in wireless, 3g/4g, etc lead to a full compromise of the device? These components don’t have full access to the device to begin with, and definitely don’t have access to the disk encryption keys.
And yeah, you can’t clone an iPhone and get anything usable. The pin is entangled with a secret that never leaves the Secure Enclave, so an offline attack would be an attack on the full encryption key, not on the pin.
They demonstrated owning an Android device. The page says iPhones since iPhone 4 use the Broadcom WiFi SoC but otherwise makes no mention of whether the exploit works on iPhone.
Most mobile phones allow the baseband to have full access to the entire device, which is why an exploit of the baseband can turn into a full device compromise, but AIUI iPhones don't do this and keep the baseband as a separate unprivileged component, specifically to defend against this attack vector. So a baseband compromise might get you access to any data going over the baseband (e.g. phone calls, unencrypted data traffic, etc) but shouldn't get you access to the rest of the device.
The radio interfaces do not have total access to the device but they have enough that it is feasible to compromise a device via a compromise of a radio component.
I do not think that a radio interface could have enough access to facilitate decryption of an encrypted volume. What I imagine it has enough access to do is to pivot to the OS running on the main CPU via a bug in the interface that is exposed for the radio to communicate with the main CPU.
From there they would likely have to exploit a number of other bugs to get into the position that they want to be in.
To protect the device from vulnerabilities in network processor firmware, network interfaces including Wi-Fi and baseband have limited access to application processor memory. When USB or SDIO is used to interface with the network processor, the network processor can’t initiate Direct Memory Access (DMA) transactions to the application processor. When PCIe is used, each network processor is on its own isolated PCIe bus. An IOMMU on each PCIe bus limits the network processor’s DMA access to pages of memory containing its network packets or control structures.
Of course there's always the chance that there's a bug in one of these interfaces
And on Android: ¯\_(ツ)_/¯ Even for the Google Pixel I can't find a security whitepaper, just a blog post with a couple of vague bullet points.
Comments
How would an exploit in wireless, 3g/4g, etc lead to a full compromise of the device? These components don’t have full access to the device to begin with, and definitely don’t have access to the disk encryption keys.
And yeah, you can’t clone an iPhone and get anything usable. The pin is entangled with a secret that never leaves the Secure Enclave, so an offline attack would be an attack on the full encryption key, not on the pin.
Google Project Zero was able to remotely own devices via a Broadcom vulnerability: https://googleprojectzero.blogspot.com/2017/04/over-air-expl...
They demonstrated owning an Android device. The page says iPhones since iPhone 4 use the Broadcom WiFi SoC but otherwise makes no mention of whether the exploit works on iPhone.
Most mobile phones allow the baseband to have full access to the entire device, which is why an exploit of the baseband can turn into a full device compromise, but AIUI iPhones don't do this and keep the baseband as a separate unprivileged component, specifically to defend against this attack vector. So a baseband compromise might get you access to any data going over the baseband (e.g. phone calls, unencrypted data traffic, etc) but shouldn't get you access to the rest of the device.
The radio interfaces do not have total access to the device but they have enough that it is feasible to compromise a device via a compromise of a radio component.
Can you be more specific about the nature of the exploit you are imagining?
For example, how can a radio interface have "enough" access to facilitate decryption of an encrypted volume?
I do not think that a radio interface could have enough access to facilitate decryption of an encrypted volume. What I imagine it has enough access to do is to pivot to the OS running on the main CPU via a bug in the interface that is exposed for the radio to communicate with the main CPU.
From there they would likely have to exploit a number of other bugs to get into the position that they want to be in.
The radios quite probably use DMA to blast bits to/from main memory.
Get code execution on the radio chip, use that to harvest the decryption keys from RAM and the rest is pretty trivial.
From the iOS security whitepaper:
Of course there's always the chance that there's a bug in one of these interfaces
And on Android: ¯\_(ツ)_/¯ Even for the Google Pixel I can't find a security whitepaper, just a blog post with a couple of vague bullet points.
Isn't the key in a part of RAM only accessible through the Secure Enclave? (As in, you shouldn't be able to set up DMA to certain ranges.)