Traffic from private or unroutable addresses

From: Russell Fulton (r.fultonat_private)
Date: Mon Jun 25 2001 - 17:08:33 PDT

  • Next message: Bjorn Djupvik: "Re: Threat mail from russia (followup)"

    Every few months we see posts to this list about traffic from various 
    private or unroutable addresses, e.g. 10.0.0.0/8 and 192.168.0.0/16.
    
    Here is an example of non malicious traffic from such a source.
    
    I'll use argus output to tell this story...
    
    Today I picked up what appeared to be a host scan from 10.0.0.107:
    
    bin/ra -AIncr data/2001.06.25/argus-2001.06.25.14.00.gz -Zb host  10.0.0.107 
    25 Jun 01 14:14:28           tcp      10.0.0.107.80     ?>    130.216.191.46.1632  5        0         7300         0           A_
    25 Jun 01 14:14:30           tcp      10.0.0.107.80     ?>    130.216.191.46.1653  5        0         7300         0           A_
    25 Jun 01 14:14:31           tcp      10.0.0.107.80     ?>    130.216.191.46.1690  5        0         7300         0           A_
    25 Jun 01 14:14:33           tcp      10.0.0.107.80     ?>    130.216.191.46.1748  5        0         7300         0           A_
    25 Jun 01 14:14:34           tcp      10.0.0.107.80     ?>    130.216.191.46.1772  5        0         7300         0           A_
    25 Jun 01 14:14:36           tcp      10.0.0.107.80     ?>    130.216.191.46.1807  5        0         7300         0           A_
    25 Jun 01 14:14:38           tcp      10.0.0.107.80     ?>    130.216.191.46.1862  5        0         7300         0           A_
    25 Jun 01 14:14:40           tcp      10.0.0.107.80     ?>    130.216.191.46.1927  5        0         7300         0           A_
    25 Jun 01 14:14:42           tcp      10.0.0.107.80     ?>    130.216.191.46.1957  5        0         7300         0           A_
    25 Jun 01 14:14:45           tcp      10.0.0.107.80     ?>    130.216.191.46.2016  5        0         7300         0           A_
    25 Jun 01 14:14:49           tcp      10.0.0.107.80     ?>    130.216.191.46.2087  5        0         7300         0           A_
    25 Jun 01 14:14:50           tcp      10.0.0.107.80     ?>    130.216.191.46.2111  5        0         7300         0           A_
    25 Jun 01 14:14:55           tcp      10.0.0.107.80     ?>    130.216.191.46.2186  5        0         7300         0           A_
    25 Jun 01 14:14:56           tcp      10.0.0.107.80     ?>    130.216.191.46.2213  5        0         7300         0           A_
    25 Jun 01 14:14:57           tcp      10.0.0.107.80     ?>    130.216.191.46.2237  5        0         7300         0           A_
    25 Jun 01 14:15:06           tcp      10.0.0.107.80     ?>    130.216.191.46.2448  5        0         7300         0           A_
    25 Jun 01 14:15:07           tcp      10.0.0.107.80     ?>    130.216.191.46.2468  5        0         7300         0           A_
    
    But why would some probe ports with ACKs (A_ means that source sent a
    packet that had ACK set and no other flags, in this case there were 5
    packets with a total of 7300 bytes of tcp data -- no replies).
    
    lets look at all traffic to or from 130.216.191.46 (a large web proxy)
    on a couple of these ports:
    
    bin/ra -AIncr data/2001.06.25/argus-2001.06.25.14.00.gz -Zb host  130.216.191.46 and port 2016
    25 Jun 01 14:14:41           tcp  130.216.191.46.2016   ->      64.47.39.100.80    9        11        307          9187        FSRPA_FSRPA
    25 Jun 01 14:14:45           tcp      10.0.0.107.80     ?>    130.216.191.46.2016  5        0         7300         0           A_
    
    bin/ra -AIncr data/2001.06.25/argus-2001.06.25.14.00.gz -Zb host  130.216.191.46 and port 2087
    25 Jun 01 14:14:45           tcp  130.216.191.46.2087   ->      64.47.39.100.80    9        10        307          7300        FSRPA_FSRPA
    25 Jun 01 14:14:49           tcp      10.0.0.107.80     ?>    130.216.191.46.2087  5        0         7300         0           A_
    
    Hmm.... in both case we see an outgoing tcp session to port 80 on
    64.47.39.100 which terminated normally (F in the status field
    indicates that FINs were exchanged). Followed 4 seconds later by
    a stream of incoming ACKs to the original source port from
    the an address in 10.0.0.0/8.
    
    I have seen many variations on this theme where sessions to a web
    server appear to terminate normally and then (up to 5 minutes later) we
    see either RSTs or ACK coming in from some other address to the original
    source port on the host that originated the session.
    
    My guess is that there are a whole lot of different causes for
    these problems and many different products involved. What they all
    have in common is that they result from imperfect systems that break
    the original end-to-end connectivity IP model.  Such devices include
    NAT enabled firewalls and load balancers.
    
    Russell Fulton, Computer and Network Security Officer
    The University of Auckland,  New Zealand
    
    
    This list is provided by the SecurityFocus ARIS analyzer service.
    For more information on this free incident handling, management 
    and tracking system please see:
    
    http://aris.securityfocus.com
    



    This archive was generated by hypermail 2b30 : Tue Jun 26 2001 - 10:50:24 PDT