Conversation

twitter.com/DanielMicay/st Android 11 disables usage of stable "privacy" addresses for networks where MAC randomization is being used. We'll no longer need to disable this feature. The other issues we've discovered and put initial work into addressing are still problems though.

Quote Tweet

Daniel Micay

@DanielMicay

Jun 24, 2020

From the latest GrapheneOS release notes (grapheneos.org/releases#2020.): > disable RFC 7217 support (stable link-local IPv6 privacy addresses) and stick to link-local IP addresses based on the (random) MAC addresses The Linux implementation of privacy extensions is some nasty stuff.

Show this thread

Daniel Micay

@DanielMicay

Sep 16, 2020

twitter.com/DontDisableIPv Silly bot. Disabling IPv6 would certainly be an easy way to address multiple issues in the Linux kernel implementation providing a way to track users across networks. Linux kernel's take on "privacy" addresses regresses privacy in some very serious ways.

Quote Tweet

Don't Disable IPv6

@DontDisableIPv6

Sep 16, 2020

Replying to @DanielMicay

Please don't disable IPv6, it will break things. #DontDisableIPv6 superuser.com/questions/1229

Daniel Micay

@DanielMicay

Linux reuses "privacy" addresses across networks. This leaks that it's the same device connecting from a different network. It does this even when randomizing the MAC address. It also uses a very long lifetime for these addresses by default. It's a pretty bad implementation.

12:11 PM · Sep 16, 2020Twitter Web App

Replying to

There are other issues beyond this too, including problematic timers and overly coarse (or simply poor) randomization. Many of the harder issues to solve are due to having the OS split up into a bunch of independent projects. Serious lack of whole picture privacy/security work.

Daniel Micay

@DanielMicay

Sep 16, 2020

If standards recommend an implementation that would require coordination between wpa_supplicant and the Linux kernel or something similar, they tend to take shortcuts resulting in weaker privacy than expected. A lot of these things have been considered in newer RFCs...

Replying to

Basically also described in several papers from 2016 or even older: papers.mathyvanhoef.com/asiaccs2016.pdf + tools.ietf.org/html/rfc7721 + (from 2014/2015) publications.sba-research.org/publications/U

Replying to

Yeah, see grapheneos.org/usage#wifi-pri. A lot of issues have been addressed over the years. Qualcomm improved Atheros firmware and drivers to support anonymity. Wi-Fi tends to leak other hardware identifiers without a way to prevent it all from the OS.

Show replies