Hello,
A couple of cuts from tcpdump output:
21:31:54.995170 IP 141.138.131.115.80 > 109.72.248.114.21: Flags [S], seq 1376379765, win 8192, length 0
21:31:55.231925 IP 194.73.173.154.80 > 109.72.241.198.21: Flags [S], seq 2254756684, win 8192, length 0
21:27:50.413927 IP 95.131.188.179.80 > 109.72.248.114.21: Flags [S], seq 3619475318, win 8192, length 0
21:27:50.477014 IP 95.131.191.77.80 > 109.72.248.114.21: Flags [S], seq 2412690982, win 8192, length 0
Does anyone seeing this right now (18:31 UTC)? I see this traffic
on at least two completely independent ISPs near Moscow. The
rate is about a few dozen PPS hitting all BGP-announced networks.
And the same sources:
141.138.128.0 - 141.138.135.255
194.73.173.0 - 194.73.173.127
95.131.184.0 - 95.131.191.255
Massive DDoS against William Hill Organization Ltd?
seeing an awful lot of port 80 hitting port 21. (Why would port 80
ever be used as source?). Also saw a buncha cpanel "FAILED: FTP" alerts flickering
on and off as the service throttled itself at a couple client sites I manage.
I see 540 unique source IPs hitting 32 destinations on my network in just 1000
packets dumped on one router.
All from multiple sequential registered /24s in whois, but all from one
management company:
141.138.128.0/21 and 95.131.184.0/21
role: William Hill Network Services
abuse-mailbox: networkservices@williamhill.co.uk
address: Infrastructure Services 2 City Walk Sweet Street Leeds LS11 9AR
AS49061
course, synfloods can be spoofed... perhaps they're hoping for a retaliation
against WHNS.
/kc
I think Ken has nailed it. I think the source addresses are spoofed so you reflect the connection (tcp syn ack) to those source addresses. Get enough of those connections and the server is dead.
Since your port 21 is open
telnet 109.72.248.114 21
Trying 109.72.248.114...
Connected to 109.72.248.114.
Escape character is '^]'.
Your address was probably scanned and saw it could be used in the attack.
Regards
Not sure why reflected RSTs are the goal here, they're not much of an amplification
to the original syn size. Additionally causing a mild dos of my clients' stuff
when it begins throttling # of connections, ie noticeable. (not that i want to
help scriptkids improve their attacks...). Im guessing port 80 was chosen for improved
fw piercing.
Sure is widespread though, 5 clients on very different networks all seeing similar
saturation. Someone has a nice complete prescanned list of open ftps for the
entire internet out there (or are they just saturating the whole /0?)
Easy to filter though:
tcp and src port 80 and src net '(141.138.128.0/21 or 95.131.184.0/21)' and dst port 21
Adapt for your fw rules of choice.
/kc
Ditto. Same sources; 141.138.128.0/21 and 95.131.184.0/21 (give or take).
Out of 1000 packet sample taken at 12:45:46 PDT (19:45:46 UTC) at boundary, 502 unique sources to 10 destination hosts on our AS.
Obligatory data should this be of use to anyone listening in.
Does the synflood have tcp option headers?
I am seeing this same activity at our forward observation system, however
it's not showing any tcp options like mss,sack,timestamps etc, was curious
if others were seeing the same
[root@oakridge-intercept(~)]> tcpdump -nn -i eth0 'tcp and (tcp[13] == 2)'
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
13:09:32.772506 IP 95.131.190.214.80 > 67.220.207.169.21: Flags [S], seq
3599006989, win 8192, length 0
13:09:32.809446 IP 95.131.185.150.80 > 67.220.207.169.21: Flags [S], seq
2409909072, win 8192, length 0
13:09:33.306737 IP 141.138.133.161.80 > 67.220.207.169.21: Flags [S], seq
1006681302, win 8192, length 0
13:09:33.946427 IP 141.138.134.193.80 > 67.220.207.170.21: Flags [S], seq
3627295948, win 8192, length 0
13:09:33.946469 IP 141.138.134.193.80 > 67.220.207.170.21: Flags [S], seq
3627295948, win 8192, length 0
13:09:34.263905 IP 194.73.173.103.80 > 67.220.207.170.21: Flags [S], seq
3818041920, win 8192, length 0
13:09:34.415558 IP 194.73.173.243.80 > 67.220.207.169.21: Flags [S], seq
3584410928, win 8192, length 0
Does the synflood have tcp option headers?
Not seeing any here. From this morning.
12:45:46.180665 194.73.173.17.80 > 216.57.181.189.21: S [tcp sum ok] 1158156467:1158156467(0) win 8192 (DF) (ttl 60, id 18499, len 40)
12:45:46.180667 194.73.173.17.80 > 216.57.181.189.21: S [tcp sum ok] 1158156467:1158156467(0) win 8192 (DF) (ttl 60, id 18499, len 40)
12:45:46.284617 141.138.128.137.80 > 216.57.182.18.21: S [tcp sum ok] 2595766696:2595766696(0) win 8192 (DF) (ttl 69, id 6478, len 40)
yeah it looks like the person behind the flood may have scanned for active
ftp servers, not seeing any activity on other observation subnets of this
flood, and so far the only servers showing this port 80 to port 21 is ones
that do have actual ftp servers, however, the connection is not actually
establishing it's only showing SYN incoming and a SYN-ACK outgoing and
never gets a completed 3way handshake, so it could be a very odd reflected
syn-ack flood against possible web servers origin ip addresses.
Yeah it is an odd ball attack for sure, here is a 5000 packet sample of
what I was seeing in connection to this attack
https://mystagic.io/80to21.pcap , don't think it's the entire /0 for ftp
port as I am not seeing it on many other subnets, which is why I am
thinking someone did a pre-scan before conducting this wacky attack,
otherwise, I would have likely seen other port 21's seeing activity, but so
far any IP that didn't have 21 as an actual service isn't seeing the syn
packets. This could be unique to my location, others observing this attack
may be able to chime in and report what they are seeing if they seen 80 src
syn to port 21 where 21 isn't an actual ftp running. Yeah this is pretty
easy to filter.
what's the density of open port 21s on the planet though? trying to estimate
the traffic resulting against the two target /21s.
Your dump only has 2 ip's in it though, on your /19 so not representative.
My dump is 500 synacks returned in 14 seconds to 32 ips in a /22. This would give
128M ftp responders across the whole /0 (modulo actual space in use, etc,
so call it 32M responders?). (It's also a short timespan for a dump as well.)
Syn-ack seems to be a 58 byte packet (?ish).
32 * 10^6 * 500/14 * 58*8 / 10^9 = 530 Gbps
even if im off by 4 in density of ftp sites on the internet despite my already
reducing it by 4, we're talking ~100+ Gbps.
/kc
Most of those networks are served by Prolexic DDOS mitigation (AS 32787),
and according to BGPlay have been for a while. (AS carrying untoward material,
like a Tor exit node or onion router?)
But a couple /24s in the 95.* block are AS14537 Mohawk Internet Tech. in
Quebec Canada such as 95.131.188.0/24 - unintended target? (careful who you
buy /24's from!)
So the only target being affected would be Mohawk unless they're setup to handle it.
/kc
There is some nice research regarding systems "abusable" for reflection by
tcp port and the amplification factor depending on the OS:
http://www.christian-rossow.de/publications/tcpamplification-woot2014.pdf
And in more detail:
https://www.usenix.org/system/files/conference/usenixsecurity14/sec14-paper-
kuhrer.pdf
Best regards,
Chris
This might be a little late on this thread, however I just saw the
following news item on twitter which seemed pertinent to this story:
http://www.theregister.co.uk/2016/11/02/william_hill_ddos/
I guess they're a bookie who's under DDoS?
Theodore Baschak - AS395089 - Hextet Systems
https://ciscodude.net/ - https://hextet.systems/
http://mbix.ca/