Yes, thanks -- that covers what I wrote.  Thanks!

I have also noticed concern with HTTP headers disclosing client-identifying 
information so might want to also mention

  "HTTP headers (e.g., User-Agent, Accept, Accept-Encoding)"

or similar.

-d

On Oct 29, 2019, at 5:47 AM, Sara Dickinson <[email protected]> wrote:
>> On 24 Oct 2019, at 01:04, Dan Wing <[email protected]> wrote:
>> 
>> I like draft-ietf-dprive-bcp-op overall.  A few comments related to 
>> fingerprinting:
>> 
>> 1. In https://tools.ietf.org/html/draft-ietf-dprive-bcp-op-04#section-5.2.4 
>> it mentions fingerprinting client OS by IPv4 TTL or IPv6 Hoplimit and TLS 
>> parameters.  The protocol between those, TCP, isn't listed, but TCP 
>> parameters (window size, ECN support, SACK) are also reasonably unique to an 
>> operating system or the application that initiated the TCP connection.  
>> 
>> 2. What does 'tracking of TCP sessions' mean in this context?
>> 
>> 3. OS fingerprinting is also possible via the DNS queries themselves, 
>> especially easy if the OS has built-in captive portal and 
>> Internet-connectivity detection mechanisms.  For example, Apple iOS 13.1 
>> queries both http://captive.apple.com/hotspot-detect.html and 
>> http://netcts.cdn-apple.com, Windows 10 queries 
>> http://www.msftncsi.com/ncsi.txt, and not sure about Android.  iOS 13.1 and 
>> Windows 10 query DNS names unique to those purposes, so far as I am aware.
>> 
>> OSs also fingerprint themselves by periodic checking for new OS versions 
>> which is usually another unique DNS query (MacOS queries swscan.apple.com 
>> whereas iOS appears to use su.itunes.apple.com, not sure about Windows; 
>> Linux would be fingerprintable down to the distribution).
>> 
>> Should the I-D recommend discarding that DNS correlation data?
>> 
>> 4. A user's machine can also be fingerprinted based on its DNS queries when 
>> it joins a network (IMAP accounts, instant messaging accounts) and its 
>> periodic 3rd and 1st party software update checks, but I guess that is sort 
>> of covered by the reference to RFC6973's Surveillance in Section 5.3?
> 
> Hi Dan, 
> 
> Thanks very much for these points. Does rewording the section as below better 
> cover this?
> 
> "DNS Privacy Threats:
> 
> *  Fingerprinting of the client OS via various means including: IP 
> TTL/Hoplimit, TCP parameters (e.g. window size, ECN support, SACK), OS 
> specific DNS query patterns (e.g. for network connectivity, captive portal 
> detection or OS specific updates).
> 
> *  Fingerprinting of the client application or TLS library by e.g. TLS 
> version/Cipher suite combinations or other connection parameters.
> 
> *  Correlation of queries on multiple TCP session originating from the same 
> IP address
> 
> *  Correlating of queries on multiple TLS sessions originating from the same 
> client, including via session resumption mechanisms"
> 
> Best regards
> 
> Sara. 

_______________________________________________
dns-privacy mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dns-privacy

Reply via email to