5 ms·

The reality is that Apple has little ability to block apps that are tracking people by indirect data such as “device boottime.” Sure maybe in that specific case

The reality is that Apple has little ability to block apps that are tracking people by indirect data such as “device boottime.” Sure maybe in that specific case they can block that piece of information but there are likely at least 100 other seemingly innocuous pieces of information available to apps that can be used the same way. The idea of a “trustless” App Store is nice and roughly works in practice but is ultimately a fiction. Using a malicious app will always expose you to danger. New vulnerabilities are announced regularly. The situation is worse on the web: WebKit vulns are released nearly monthly and there is no app review there. Browsing the web with JavaScript enabled is a security nightmare. For example see the recent security notes for iOS 14.4.2 [1], every release you’ll likely see a fix for a new vulnerability. I’m hesitant to say there will ever be a scalable and permanent solution to this problem. The best advice is simple: do not run apps you do not trust, do not visit websites you do not trust. [1] https://support.apple.com/en-us/HT212256 https://support.apple.com/en-us/HT212256

pornel 6y ago
Apps are heavily sandboxed. iOS controls all local information they can access, so Apple is in position to fight this. It doesn't need to be perfect. Only hard enough that apps have to resort to more desperate techniques, so that they're detectable and Apple can ban them for trying. Overall I don't like Apple's sandbox and total control over iOS, but it exists, and in this case at least they can use their power for something good.
- CyberRabbi 6y ago
  The point of the article is exactly that it isn’t 100% perfect and this slight imperfection has allowed companies like P&G to take advantage of that for large-scale privacy violation. From the article: The whack-a-mole1 aspect of Apple’s new privacy rules is that while Apple can restrict access to the API that provides access to the IDFA identifier, clever developers can find (perhaps infinite) other ways to combine things they do have access to into a unique, or even just “close enough to unique to be useful for tracking”, identifier. IP addresses, to name just one example, are a big factor that Apple can’t block would-be-trackers from using.
buran77 6y ago
All apps rely on information provided to them by the OS. The OS can simply introduce (pseudo)random variability in that data, especially for apps that collect a lot of it so the probability for aggregating and correlating is a lot higher. The more access your app wants, the more randomness should be introduced.
- CyberRabbi 6y ago
  What about the IP address used to communicate with the app’s server? Can that be obscured? What about the screen resolution? What about time zone? Battery level?
  - notriddle 6y ago
    Screen resolution? You can predict that perfectly based on the model number, and many models have the same screen resolution. https://developer.apple.com/library/archive/documentation/DeviceInformation/Reference/iOSDeviceCompatibility/Displays/Displays.html https://developer.apple.com/library/archive/documentation/De...
  - tinus_hn 6y ago
    Nobody’s blocking IP addresses but apparently these are not enough or these companies wouldn’t be clutching their other tracking pearls. IP addresses and battery levels are dynamic and screen resolution and time zones are shared by millions. The real fingerprinting is in things like sensor calibration and these can be stopped by binning or randomizing values. Apple has announced they will be blocking these prints one by one and they are. Developers wouldn’t be screaming if they weren’t in trouble.
    - CyberRabbi 6y ago
      The point of the article is that there are enough data that are low information in isolation but when combined together can effectively identify ad targets. This is already being done, it isn’t a theory. That’s the point of the article: clever developers can find (perhaps infinite) other ways to combine things they do have access to into a unique, or even just “close enough to unique to be useful for tracking”, identifier
Nextgrid 6y ago
This is why tracking is a problem that can only be solved at the legal level. Ultimately there's no reliable way for a machine to detect tracking given that any information collected for tracking can also be used for legitimate purposes.
- CyberRabbi 6y ago
  Just like App Store review, it can only be mitigated at the legal level. Companies will skirt the law if there is little risk of them being caught and, like you said, there will never be a robust way of automatically detecting violations.

pornel 6y ago

Apps are heavily sandboxed. iOS controls all local information they can access, so Apple is in position to fight this. It doesn't need to be perfect. Only hard enough that apps have to resort to more desperate techniques, so that they're detectable and Apple can ban them for trying. Overall I don't like Apple's sandbox and total control over iOS, but it exists, and in this case at least they can use their power for something good.

CyberRabbi 6y ago
The point of the article is exactly that it isn’t 100% perfect and this slight imperfection has allowed companies like P&G to take advantage of that for large-scale privacy violation. From the article: The whack-a-mole1 aspect of Apple’s new privacy rules is that while Apple can restrict access to the API that provides access to the IDFA identifier, clever developers can find (perhaps infinite) other ways to combine things they do have access to into a unique, or even just “close enough to unique to be useful for tracking”, identifier. IP addresses, to name just one example, are a big factor that Apple can’t block would-be-trackers from using.

buran77 6y ago

All apps rely on information provided to them by the OS. The OS can simply introduce (pseudo)random variability in that data, especially for apps that collect a lot of it so the probability for aggregating and correlating is a lot higher. The more access your app wants, the more randomness should be introduced.

CyberRabbi 6y ago
What about the IP address used to communicate with the app’s server? Can that be obscured? What about the screen resolution? What about time zone? Battery level?
- notriddle 6y ago
  Screen resolution? You can predict that perfectly based on the model number, and many models have the same screen resolution. https://developer.apple.com/library/archive/documentation/DeviceInformation/Reference/iOSDeviceCompatibility/Displays/Displays.html https://developer.apple.com/library/archive/documentation/De...
- tinus_hn 6y ago
  Nobody’s blocking IP addresses but apparently these are not enough or these companies wouldn’t be clutching their other tracking pearls. IP addresses and battery levels are dynamic and screen resolution and time zones are shared by millions. The real fingerprinting is in things like sensor calibration and these can be stopped by binning or randomizing values. Apple has announced they will be blocking these prints one by one and they are. Developers wouldn’t be screaming if they weren’t in trouble.
  - CyberRabbi 6y ago
    The point of the article is that there are enough data that are low information in isolation but when combined together can effectively identify ad targets. This is already being done, it isn’t a theory. That’s the point of the article: clever developers can find (perhaps infinite) other ways to combine things they do have access to into a unique, or even just “close enough to unique to be useful for tracking”, identifier

Nextgrid 6y ago

This is why tracking is a problem that can only be solved at the legal level. Ultimately there's no reliable way for a machine to detect tracking given that any information collected for tracking can also be used for legitimate purposes.

CyberRabbi 6y ago
Just like App Store review, it can only be mitigated at the legal level. Companies will skirt the law if there is little risk of them being caught and, like you said, there will never be a robust way of automatically detecting violations.