SYN-ACK Packets With Data

Published: 2016-01-27
Last Updated: 2016-01-27 19:06:07 UTC
by Johannes Ullrich (Version: 1)
1 comment(s)

We haven't had an event like this in a while... "Odd Packets"! I was going through some honeypot packet captures with tcpflow, when I got this error message:

$ tcpflow -r ../allpackets
tcpflow: TCP PROTOCOL VIOLATION: SYN with data! (length=970)

It has been a while since I got SYN packets with data! So I had to look closer:

$ tcpdump -r ../allpackets -nX 'tcp[13]=2 && ip[2:2]-
reading from file ../allpackets, link-type EN10MB (Ethernet)

Nothing! Is tcpflow wrong? Well... I may be a bit too picky with tcp[13]=2. Lets make Judy proud and use a bitmask:

tcpdump -r /tmp/allpackets -xn 'tcp[13]&2=2 && ip[2:2]-
reading from file /tmp/anon2, link-type EN10MB (Ethernet)
08:43:59.138235 IP > Flags [S.],
seq 159625496:159626466, ack 770903892, win 12960, length 970
    0x0000:  4508 03f2 530f 4000 2e06 71eb c000 0201  
    0x0010:  c000 0202 2382 6b3a 0983 b118 2df3 0f54  
    0x0020:  5012 32a0 6ec5 0000 0000 0000 0000 0000  
    0x0030:  0000 0000 0000 0000 0000 0000 0000 0000  

Here we got it. It was actually a SYN-ACK, not a SYN that had the payload. The payload was all 0x00 (I truncated the output).

There was no SYN going to that IP address, so this was an unsolicited response ("backscatter"). Has anybody seen traffic like this? So far, this was the only packet I have seen. The original source IP was DoS agains the analyst? Or some kind of new TCP based reflective DoS off a real broken service?

Johannes B. Ullrich, Ph.D.

1 comment(s)


Maybe it’s just TCP Fast Open? Apparenly that’s been on by default in Linux since January 2014. I guess it can’t be considered a TCP protocol violation any more.

Using TFO to send a bunch of zeros doesn’t seem to make much sense for any legitimate use, though.

For anyone who wants to catch payload-carrying SYNs with iptables, there’s a nifty trick at the bottom of this page:

Diary Archives