On 02/14/2014 09:02 AM, Giles Davis wrote:
> Nick Hilliard wrote:
>> It really is, but bear in mind that a single 1GE connection with
>> no urpf can be used to create ~250-300G of backscatter traffic.
>> 
>> This means that there's only a requirement to have a single 
>> unscrupulous or incompetent ISP with GE in the world to allow a 
>> devastating DoS to be launched against anyone anywhere.
>> 
> Indeed - which is certainly a problem! :)
> 
> So what's the 'proper' solution to all this then beyond just adding 
> enough capacity to absorb ever larger attacks? How's this going to 
> end up?

What's happening now is that reactive, specific measures are being taken
- protocol-specific vulnerabilities (e.g. RRL for DNS, disable monlist
for NTP) are being plugged, ISPs are deploying better instrumentation to
detect attack flows, and are turning on uRPF/other source-address
filtering towards the worst traffic sources.

The problem with these approaches is that:
- they are just going to lead to an endless game of whack-a-mole as the
  bad guys find ever more reflection vectors which need plugging
- this arms race will in turn educate the bad guys to be smarter
- the vendors of security products are going to be more interested in
  selling bigger faster $olutions than tackling the underlying problems
  (cf ever-increasing claims for how big an attack various vendors
   claim to have dealt with)
- TPTB are more likely to blame the Internet industry and take
  regulatory measures against us as an easy target than tackle the
  actual bad guys

Something I think that would make a bigger difference would be for data
to be gathered and published that names-and-shames those providers that
don't do BCP38 source address validation. As an industry we then need to
start contractually enforcing, de-peering and blocking traffic to/from
those providers who don't take action to remedy this.

The other thing is to beat up on our vendors - I hear many stories of
how BCP38 cannot be implemented by people who want to, due to some bug
or missing feature with CPE/edge/aggregation/core equipment.

If self-regulation doesn't work, we can expect regulation. While
mandating SAV/BCP38 would IMHO be a much more useful single item of
legislation to reduce Internet evil than the swathes of vested-interest
pandering nonsense we've had from our governments and regulators lately,
it's hard to trust them to do it right.

In any case, there are also many (probably most) nation-states out there
proudly declaring that they have "cyberwarfare capability", and it's
hard to see how this is credible without a DDoS element. It might
actually take international "Internet disarmament" treaties to nail this
problem :-(

Keith

Reply via email to