[arch-d] Re: IP Address Geolocation Considered Harmful

[Christian Huitema <[e-mail]>]

On 3/23/2026 1:23 PM, Brian E Carpenter wrote:
> On 24-Mar-26 06:44, Phillip Hallam-Baker wrote:
>>
>>
>> On Mon, Mar 23, 2026 at 1:31 PM Tony Li <[e-mail] 
>> <mailto:[e-mail]>> wrote:
>>
>>
>>
>>      > On Mar 23, 2026, at 9:45 AM, John Levine - johnl at ietf.email 
>> <[e-mail] <mailto:[e-mail]>> 
>> wrote:
>>      >
>>      > How do you plan to do routing without some sort of geographic 
>> aggregation?
>>
>>
>>     Routing depends on topological aggregation, not geographic 
>> aggregation.
>>
>>
>> Agree but I think we need to go a bit further as what people are 
>> doing currently is to use the habits of ISPs in domain allocation to 
>> infer geolocation.
>
> I assume that "they" are doing more than that. Knowing that routing is 
> topological, "they" can use longest match against known targets to 
> deduce the likely location of the unknown target. I happen to be in a 
> hotel in $CITY today, and I just checked where a geolocation company 
> thinks I am (based on the first traceroute hop outside the hotel's 
> NAT), and they got $CITY right as well as the correct postal code.

"They" are persistent and sophisticated attackers of the Internet. They 
are not looking just at the IP address prefix. They are also correlating 
IP addresses with data from cookies, advertisement auctions, payment 
systems and the like. I was once told that they only need to advertise 
data from less than 3 dozen web transactions to build a fairly accurate 
profile of a new user, and they probably need even less than that to 
identify the location of an IP address. On the other hand, despite all 
the data leaking all over the place, geolocation companies appear to get 
it wrong quite often.

>
> On the other hand, I doubt that a recommendation against doing this 
> would make such companies blink an eye, and law enforcement agencies 
> will do it anyway.


Of course they won't. On the other hand, there are things that the IAB 
and the IETF can do to discourage this malevolent practice. 
Standardizing VPNs and other privacy technologies, encouraging network 
providers to aggressively renumber their consumer networks, repeating 
the "address depend on topology. not geography", stating that from that 
point of view the disconnection between satellite connectivity and 
geography is perfectly fine. I am sure I am missing some.

-- Christian Huitema