Re: Recent Attacks

From: Bennett Todd (betat_private)
Date: Fri Feb 18 2000 - 07:46:05 PST

  • Next message: Ryan Russell: "Re: Recent Attacks"

    --oFbHfjnMgUMsrGjO
    Content-Type: text/plain; charset=us-ascii
    
    Sounds like you have a nice response strategy for many attacks.
    But....
    
    2000-02-16-18:42:18 Barrett G. Lyon:
    > I first identify what type of attack in coming in and what the
    > targets of the attack are (host/service/user/..etc..).  I then
    > immediately drop and log all packets from attacking network(s) (to
    > reduce load on attacked machines if it is a SYN attack, etc). [
    > granted it is not possible with some sorts of attacks to drop and
    > log everything ]
    
    This is one of those sorts. Random source addresses, random ports,
    you can disconnect yourself from the net, but why bother, the
    attacker has already done that for you.
    
    The first hint you have something is going wrong looks more like a
    router crash than an attack. At least according to SANS, Yahoo was
    getting hit with as much as a gigabyte per second of data. I don't
    know about you, but if someone pointed something that big at me, the
    result wouldn't look like an attack on _me_, it would completely
    nuke my provider and all their customers, blow us all clean off the
    air.
    
    > You need to figure out who is actually doing the attack and
    > notify their providers with a clean description of what actually
    > took place. [...] When I think about it there is also an entire
    > forensics process of figuring out who was/is doing the attack.
    > I've found that before an attack begins the attacker usually does
    > a port scan or some sort of survey of the services on the target
    > system and usually the attacker does this from their own host and
    > not another host.  They figure you will never link the attack to a
    > port scan or whatever the survey may be. . .
    
    When this DDoS is unleashed on you, you haven't been probed before;
    they aren't hunting for vulnerabilities, they're clogging your pipes
    and overwhelming your routers and servers.
    
    The packets you receive have random forged source addresses, so you
    can't track them to their source by examining the packets. The only
    thing you can tell by examining the packets is which router last
    touched that packet. So you can backtrace one step at a time, but as
    soon as you get to your borders, the backtraces start fanning out
    exponentially, since these attacks are mounted using thousands of
    systems simultaneously.
    
    It's possible that the attackers might have port-scanned the victims
    before launching these attacks --- anything's possible. But there's
    sure no reason for it; unless the attackers were dumb, the very
    first and only packets that hit the victims in any way connected
    with this attack or the attackers, were the torrents from the
    thousands of zombies, pouring in with forged source addrs.
    
    > I could write an entire book on this subject but my point is that
    > I really don't think large corporations are equipped to handle
    > nearly any type of DoS attack.  They don't understand the dynamics
    > of the attacks and they don't understand the methods of surviving
    > an attack.
    
    This one you don't survive, and you don't handle. You lie there dead
    until the attacker takes their foot off your throat. In theory,
    perhaps, if the attacker were to have tried to keep one of these
    sites down for days or weeks, it might have been possible to start
    backtracking enough of the zombie hosts to begin to thin out the
    DDoS net. And it's always possible that the attacker has left enough
    tracks at one of the zombies to make it possible to backtrack them
    completely. But I wouldn't bet on it.
    
    > Too bad it is not possible for providers to practice proper egress
    > filter techniques, because after all that is what this is all
    > about.
    
    It's not only possible, that's the only real fix. It'll happen, too.
    
    But I'm not sure it'll be primarily the providers; I think despite
    the growth of the internet providing business a thousand-fold or
    more in the 1990s, the total amount of internet design and admin
    clue has remained constant. There are a lot of morons running
    backbones these days.
    
    I suspect ingress filtering will be rolled out faster by connecting
    customers than by providers. It's trivially easy to do, and doing it
    gets you immediate benefits. While you're at it, slap down IP
    Directed Broadcast filtering, too.
    
    -Bennett
    
    P.S. I've done a witepaper on DDoS at
    <URL:https://fridge.oven.com/~bet/DDoS/>.
    
    --oFbHfjnMgUMsrGjO
    Content-Type: application/pgp-signature
    
    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.0.0 (GNU/Linux)
    Comment: For info see http://www.gnupg.org
    
    iD8DBQE4rWk9L6KAps40sTYRAfu5AJ9eUYI+H4MmtyZ/wgmYGNhlEhuIaQCgkA40
    1wxLg+lju6L1z2+7EcEsF1s=
    =oXlZ
    -----END PGP SIGNATURE-----
    
    --oFbHfjnMgUMsrGjO--
    



    This archive was generated by hypermail 2b30 : Fri Apr 13 2001 - 14:06:03 PDT