3 ms·

But it’s a good question. I want to know. I am assuming this is not possible. The only thing i know of is capable of doing so is pegasus. But it’s very expensiv

But it’s a good question. I want to know. I am assuming this is not possible. The only thing i know of is capable of doing so is pegasus. But it’s very expensive afak.

anonym29 3y ago
You don't know what code is running on your baseband processor, do you? Do you know what other hardware your baseband processor has the ability to inspect?
- zozbot234 3y ago
  In most SoC's the answer is 'everything' because there's no such thing as an IOMMU.
  - anonym29 3y ago
    Ding ding ding, we have a winner!
  - paulryanrogers 3y ago
    How big a concern is this if the data is encrypted by the kernel or user space?
    - Veserv 3y ago
      Encryption does not help in this case. They have complete remote control over the entire CPU so they can just run the decryption code directly. Encryption only helps if the endpoints that can get access to the plaintext are not compromised.
    - ignoramous 3y ago
      There are atleast 2 more exception levels with higher privileges than the Kernel on arm64.
  - moyix 3y ago
    I was under the impression that most modern (past few years) SoCs like Exynos, Qualcomm, Apple silicon all had IOMMU support. Sometimes it’s misconfigured to be too permissive but that’s getting better. Qualcomm SMMU: https://www.qualcomm.com/content/dam/qcomm-martech/dm-assets/documents/whitepaper_0.pdf https://www.qualcomm.com/content/dam/qcomm-martech/dm-assets... Apple: https://support.apple.com/lt-lt/guide/security/seca4960c2b5/web https://support.apple.com/lt-lt/guide/security/seca4960c2b5/... Samsung (vuln indicating it wasn’t configured correctly, but they still do have and use an IOMMU): https://nvd.nist.gov/vuln/detail/CVE-2022-39854 https://nvd.nist.gov/vuln/detail/CVE-2022-39854
  - numpad0 3y ago
    Why's IOMMU thrown around so casually in this forum as if it's a silver-bullet explosive reactive armors? They'd be running something like 30 years old giant main loop with "// don't remove this line, build breaks" comments everywhere, not like Rust microservices on formally verified microkernel. The main CPU/application processor/main CPU might be running better secured Unix/Linux and might be able to protect itself from peripheral CPUs, but that's not the point; a phone had always been a pair (minimum) of computers, traditionally referred to as Application Processor(AP) and Baseband Processor(BP), of only the slightly faster one is exposed to the user, and it's unclear what is going on inside the other one or how to handle it. That's the problem.
- freecodyx 3y ago
  Ok but we are talking remotely enabling camera and microphone. The baseband is only responsible of intercepting traffic. This needs kernel injection.
Veserv 3y ago
It costs about 2-5M$ to buy or develop a new weaponized zero-click vulnerability that would allow you to simultaneously hack all 1,000,000,000 iPhones in use. So around 1/20 of a cent per iPhone.
- abwizz 3y ago
  even for the single use case, 5M$ is not that far fetched in terms of opportunity cost.

But it’s a good question. I want to know. I am assuming this is not possible. The only thing i know of is capable of doing so is pegasus. But it’s very expensive afak.

anonym29 3y ago

You don't know what code is running on your baseband processor, do you? Do you know what other hardware your baseband processor has the ability to inspect?

zozbot234 3y ago
In most SoC's the answer is 'everything' because there's no such thing as an IOMMU.
- anonym29 3y ago
  Ding ding ding, we have a winner!
- paulryanrogers 3y ago
  How big a concern is this if the data is encrypted by the kernel or user space?
  - Veserv 3y ago
    Encryption does not help in this case. They have complete remote control over the entire CPU so they can just run the decryption code directly. Encryption only helps if the endpoints that can get access to the plaintext are not compromised.
  - ignoramous 3y ago
    There are atleast 2 more exception levels with higher privileges than the Kernel on arm64.
- moyix 3y ago
  I was under the impression that most modern (past few years) SoCs like Exynos, Qualcomm, Apple silicon all had IOMMU support. Sometimes it’s misconfigured to be too permissive but that’s getting better. Qualcomm SMMU: https://www.qualcomm.com/content/dam/qcomm-martech/dm-assets/documents/whitepaper_0.pdf https://www.qualcomm.com/content/dam/qcomm-martech/dm-assets... Apple: https://support.apple.com/lt-lt/guide/security/seca4960c2b5/web https://support.apple.com/lt-lt/guide/security/seca4960c2b5/... Samsung (vuln indicating it wasn’t configured correctly, but they still do have and use an IOMMU): https://nvd.nist.gov/vuln/detail/CVE-2022-39854 https://nvd.nist.gov/vuln/detail/CVE-2022-39854
- numpad0 3y ago
  Why's IOMMU thrown around so casually in this forum as if it's a silver-bullet explosive reactive armors? They'd be running something like 30 years old giant main loop with "// don't remove this line, build breaks" comments everywhere, not like Rust microservices on formally verified microkernel. The main CPU/application processor/main CPU might be running better secured Unix/Linux and might be able to protect itself from peripheral CPUs, but that's not the point; a phone had always been a pair (minimum) of computers, traditionally referred to as Application Processor(AP) and Baseband Processor(BP), of only the slightly faster one is exposed to the user, and it's unclear what is going on inside the other one or how to handle it. That's the problem.
freecodyx 3y ago
Ok but we are talking remotely enabling camera and microphone. The baseband is only responsible of intercepting traffic. This needs kernel injection.

Veserv 3y ago

It costs about 2-5M$ to buy or develop a new weaponized zero-click vulnerability that would allow you to simultaneously hack all 1,000,000,000 iPhones in use. So around 1/20 of a cent per iPhone.

abwizz 3y ago
even for the single use case, 5M$ is not that far fetched in terms of opportunity cost.