Folks,
After over 10 (yes, *ten*) years, we have finally addressed
security/privacy issues in the generation of IPv6 stable addresses in
most popular operating systems.
The traditional scheme/algorithm to generate stable IPv6 addresses with
SLAAC required that the underlying MAC address be employed to generate
the Interface Identifier. That is, the underlying MAC address would be
embedded in the lower bits of an IPv6 address.
This scheme allowed for host-tracking (since MAC addresses are usually
globally-unique), address scanning (since addresses will follow specific
patterns) and a number of other issues.
In 2011, I submitted an IETF Internet-Draft proposing a scheme for
generating stable addresses with SLAAC, meant to replace the traditional
scheme. The scheme could be summarized and simplified as: Interface_ID =
Hash(Prefix, Secret). Thus, interface identifiers would be stable within
the same subnet, but vary across subnets.
[Replacing the traditional scheme with this new scheme was anything but
easy -- if you're curious, please check the "IPv6 Addressing" section in
<https://www.si6networks.com/2020/08/06/a-brief-history-of-recent-advances-i…>
]
Over time, popular operating systems and packages adopted the proposed
algorithm: the Linux kernel, NetworkManager, OpenBSD's slaacd, MacOS,
etc. Eventually, virtually every popular OS had adopted the scheme....
except Windows.
Based on a recent note by Brian Carpenter, I ended up testing Windows
11, and I can confirm that it does implement RFC 7217 / RFC 8064!
Therefore, e.g. if multiple prefixes are employed on a subnet, the
stable addresses for each of such prefixes will employ a different
Interface Identifier, thus avoiding the security/privacy issues
discussed above -- this is really good news!
Unfortunately, Windows still generates temporary addresses with the
algorithm specified in RFC 4941, thus resulting in all temporary
addresses for a given interface employing the same Interface Identifier
(!). This problem has been addressed in RFC 8981... but it's
implementation is not yet widespread, yet (it has been incorporated in
e.g. the Linux kernel, though).
I just hope it doesn't take Windows and others yet another 10+ years to
implement RFC 8981, to finally address the remaining security/privacy
issues in IPv6 address generation!
[Original article with screenshots:
https://www.linkedin.com/posts/fernandogont_after-over-10-yes-ten-years-we-…
]
Thanks!
Regards,
--
Fernando Gont
SI6 Networks
e-mail: fgont(a)si6networks.com
PGP Fingerprint: F242 FF0E A804 AF81 EB10 2F07 7CA1 321D 663B B494
