Conversation

Ideally the Linux kernel would do something like a keyed hash of the MAC address with a random secret. Not how either stable or regular privacy addresses work. It actually *causes* privacy issues when using full MAC randomization. Love when privacy features make things worse!
1
3
We have more work to do in this area. The stable privacy address feature leads to reusing link-local addresses across different networks. The implementation makes no sense to me. It's strictly worse than always using link-local addresess based on MAC address. What's the point?
1
2
MAC address is available on the local network. If it's randomized, which it is by default in AOSP (persistent per-network random MAC) and GrapheneOS (stateless by default) then it just makes things worse. If there's no MAC randomization, it doesn't make things any better...
1
1
For public addresses, privacy addresses (regular or stable) are better than using a non-random MAC address. I think the way it's done is wrong though. If it did hmac(secret, mac_address), it wouldn't hurt. Instead it uses the secret entirely separately, screwing up privacy.
1
2
Not really sure what to do about it yet. Going to try to figure out a fix for the next release. As mentioned in earlier threads, there are deeper issues in the Linux kernel TCP/IP stack allowing users to be fingerprinted across networks until they reboot. It's hard to solve this.
2
2