Re: [saag] can an on-path attacker drop traffic?

[Dan Harkins <[e-mail]>]

On 10/3/20 10:23 PM, Christian Huitema wrote:
>> On Oct 3, 2020, at 8:00 PM, Eric Rescorla <[e-mail]> wrote:
>>
>>     The QUIC 21.13.3.1 "on-path" attacker seems to be a Dolev-Yao
>>     attacker.
>>
>>     The 21.13.3.2 "off-path" attacker seems to have the ability to
>>     observe
>>     packets, which I normally would not think an off-path attacker
>>     would have.
>>     So this definition is very surprising to me.
>>
>>
>> I agree it's not ideal. QUIC has been pubreq-ed, so you could raise 
>> it in IETF-LC.
>
> I think of it as man-in-the-middle, man-on-the-side and 
> man-in-the-rough. For me, the man in the middle is what EKR refers to 
> as the Dolev-Yao attacker; best hope there is to detect the attack and 
> reduce it to a denial of service.
>
> The man on the side is capable of reading the traffic and injecting 
> messages; it cannot easily delete messages,  but it can win races and 
> get his own fakery delivered before the genuine packets -- TCP RST is 
> an example of such attacks. Various national organizations have that 
> capability. It is much easier for them to implement than a full MITM. 
> I think that with effort we can defeat this class of attackers.
>
> The man in the rough does not see the traffic but can make guesses. 
> DDOS attacks, port scanning, observing or gaming DNS caches fall in 
> that category. Botnets sometimes resort to these attacks.

   This is useful. Defining it as classes of capabilities. The 
man-in-the-rough has
certain capabilities, class C. The man-on-the-side has those plus some 
others, class B.
And the man-in-the-middle (the Dolev-Yao attacker) has those plus some 
others, class
A. And A > B > C.

>
> And yes, in 2020 we probably need names that don't carry "man-in-foo" 
> imagery. But English is not my mother tongue...
   Yes, and sadly, this being 2020, identity politics plus bike shedding 
will mean most
of the time is spent on naming/imagery.

   Dan.