Last Updated: 2010-03-03 22:48:29 UTC
by Daniel Wesemann (Version: 1)
Following up on Mark Hofman's earlier post ("What is your firewall log telling you"), here's some tips on how a Unix command line can be used to cut a big firewall log file down to size for quick analysis.
In the installation that I use as an example here, the firewall sits between the internal network and the Internet, and only allows the internal proxy server to talk to the Internet. PC's on the internal network must surf over the proxy, and are barred from making direct connections through the firewall. This doesn't mean though that the PC's won't try -- a lot of malware, chat software, file sharing tools, etc are proxy-aware, but frequently also try a "direct" connection first. Looking at the direct connections that internal PCs attempt to make is a good way to find out what's happening on the network...
Let's assume we have a Checkpointish log format like this:
time=2010-03-02 23:59:57 action=drop orig=192.168.1.103 i/f_dir=inbound i/f_name=eth1c0 has_accounting=0 product=VPN-1 & FireWall-1 policy_name=INTERNET src=188.8.131.52 s_port=37586 dst=184.108.40.206 service=3384 proto=tcp rule=16 xlatesrc=220.127.116.11 xlatesport=57517 xlatedport=0 NAT_rulenum=4 NAT_addtnl_rulenum=internal
Let's further assume that we have already checked the connections that the firewall accepts, and that we now only want to look at traffic the firewall blocked. The easiest way to accomplish this is by suppressing all log lines that contain the string "action=accept", as follows
| grep -v "action=accept"
Now is where things get interesting. To make sense of the huge volume of log lines, we have to somehow "boil them down" into the essentials. Piping a log line into a command like this
| cut -d" " -f14-16
will just give us the fields 14-16 of each log line, with "space" used as separator (-delimiter). Your log line format might vary of course, so just count and adapt. In the above example, fields 14-16 amount to "dst=18.104.22.168 service=443 proto=tcp", so this gives us the destinations that our internal PCs attempt to talk to. Applying this "cut" command to the entire log still will give us the same number of lines as we had in the original log, but adding a command like
| sort | uniq -c | sort -rn
combines and counts all identical entries (uniq -c), and then gives us those lines that have the highest counts (sort, -n: by number, -r : reverse order). The result on the firewall log that I used as example here was
36642 dst=22.214.171.124 service=80 proto=tcp
8616 dst=126.96.36.199 service=80 proto=tcp
2641 dst=192.168.0.1 service=657 proto=udp
Once at this stage, the next step is to find out what these destinations are, and which systems we have on the network that are trying to talk to them. In this example, a quick check of "whois" reveals that both the top two IP addresses are in Russia ... and the topmost address had 36642 connection attempts from our network in a single day. Hmmm.....
Extending our original "cut" command above into
| cut -d " " -f12,14-16
now also includes the "source address" (field 12) into the output. If we now filter (grep) the output to only contain the first of the above russian IPs, the entire command line becomes
cat logfile.20100302 | grep -v "action=accept" | cut -d" " -f12,14-16 | grep "188.8.131.52" | sort | uniq -c | sort -rn
36424 src=192.168.140.108 dst=184.108.40.206 service=80 proto=tcp
218 src=192.168.143.19 dst=220.127.116.11 service=80 proto=tcp
which shows two clients on our internal network (192.168.x) that try to talk to this outside IP address.
Now, the next step is to isolate the process(es) on these two systems that initiate the connections. If the two clients run on Windows, the best bet here is to use "tcpview" from the SysInternals suite. Some patience is called for - since the connection gets dropped by our firewall, it will not be listed as an active (ESTABLISHED) connection. But watching tcpview should eventually show a process that tries to open an outbound connection ("SYN_SENT"). Once the process is known, the "What Does Your Firewall Log Tell You" portion of this investigation is over, and standard malware fighting procedures kick in. SysInternals "ProcessExplorer" is a good next step to learn more about the potentially malicious process.
What is your favorite way to quickly carve out interesting tidbits from huge firewall logs? Let us know!
Last Updated: 2010-03-03 17:28:42 UTC
by Johannes Ullrich (Version: 1)
This particular attack appears to be a new version of similar e-mails we have seen over the last week or so. The new version uses larger e-mail messages, which appear to be composed with Microsoft Word.
The text is still pretty concise. As a sample:
----- Dear Customer, Your order has been sucessfully confirmed. For your reference, here's a summary of your order: You just confirmed order #2341-23483720-38123 Status: CONFIRMED -----
At the end of the e-mail follows a link to a malware site, labeled "ORDER INFORMATION".
A number of different domains have been seen used so far.
Last Updated: 2010-03-03 10:33:22 UTC
by Mark Hofman (Version: 1)
Microsoft has re-released the patch for MS10-015 (http://blogs.technet.com/msrc/archive/2010/03/02/update-ms10-015-security-update-re-released-with-new-detection-logic.aspx). Reader Brian noticed that the patch in his WSUS had expired today and correctly surmised that an update was imminent.
A tool has also been released that will scan machines for compatibility with the new patch. This is also available from the above link.
The update will not be applied to systems on which the original patch is already installed.