sks-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Sks-devel] simple DoS against SKS's HKP interface


From: Pacal Mayan
Subject: Re: [Sks-devel] simple DoS against SKS's HKP interface
Date: Mon, 19 Mar 2012 23:22:43 -0500
User-agent: Mozilla/5.0 (X11; Linux i686; rv:10.0.2) Gecko/20120216 Thunderbird/10.0.2

On 03/18/2012 09:46 PM, John Clizbe wrote:
The default setting for wserver_timeout is 180 seconds.
Does setting it to a lower value in sksconf help?
I just tested with 10 instead of 180.

if i revert my nginx changes and allow sks back to listening on public
ports, set wserver_timeout: 10 in /etc/sks/sksconf, and restart the sks
daemons, then yes: a single malingering connection can only block the
server for 10 seconds instead of 180 seconds.

The downside of this, of course, is that bandwidth-constrained clients
(like the one i mentioned earlier as performing an accidental DoS) will
probably just fail in their connections, however, since their tcp
streams are often stuttered with pauses>= 10 seconds.

So wserver_timeout: 10 is something of an improvement over the default
of 180, but it introduces its own problems.  And the bar for a malicious
attacker isn't significantly raised -- they just need to make a new
request every 10 seconds instead of every 180 seconds, since the
settings change doesn't address the underlying concurrent requests issue).

As workarounds go, i think a reverse HTTP proxy is a better workaround,
but setting "wserver_timeout: 10" is probably a worthwhile improvement
if there's some reason that folks can't afford to just "apt-get install
nginx" (or whatever the equivalent is for your OS or your reverse proxy
of choice) immediately.

Regards,

        --dkg


would implementing an accept filter help? i.e., accf_data or accf_http on the socket?




reply via email to

[Prev in Thread] Current Thread [Next in Thread]