Intelligent IPS

April 4th, 2008

Some time ago, I have informed my colleague to report an issue that URL request to retrieve Awstats of one domain has been blocked by Intrusion Prevention System (IPS) of the Data Center, because I don’t see my request is received by web server. For example,[DOMAIN NAME] , as long as your domain name contains the word “system”, then your request is blocked.

This is their first reply (I just copied and pasted here) from the Data Center Support:

I beleive this is application issue and something to do at your end/server, please check. All your IP able to ping, we didnt block anything sort of application.

OK, fine, usual reply. We know he is one of those high position technical members, we tried to clarify that with few URLs with explanation, then his next reply:

I’ve checked already at my IPS. Your application hit one of our filter and has been blocked, pls refer below filter description

Severity : Critical
Description : This filter detects an attempt to exploit an input validation vulnerability present in the AWStats log analyzer. If successfully exploited, and attacker could execute arbitrary code on the affected web server.

Great, just as what I have suspected. So, I have replied him this is most probably a false positive, is it possible to disable the signature, or improve the signature? But I got a reply:

The IPS block a valid request, thats mean there is a vulnerability with your AWStats.

WTF? Does that mean their IPS is so intelligent and it can detect the latest Awstats version installation has a vulnerability? I have requested them to provide the full packets detail, which shows I am trying to do arbitrary code execution or maybe I can inform Awstats on the “vulnerability”.

Finally he replied:

Pls find the attachment a log from our IPS (It contains a few lines of successful URL blocking, with source and destination IP, severality, hits, that’s all). However we will consult with our vendor to determine the ’signature’ from our IPS.

I think I need to wait at least for a few weeks before get this resolved.

Sguil 0.70 and Wireshark 1.0.0 Released

March 31st, 2008

Sguil LogoIt seems a bit late to announce both releases. After a long development process and beta testing, Sguil 0.7.0 has finally been released. It does not take long for modsec2sguil to work with the latest release as well. I have not upgraded my Sguil installation to 0.70 yet, but it won’t take a long time for me to do that because I plan to upgrade the hard disk as well, so every thing will be fresh installation, including the OS - FreeBSD 7.0.

Wireshark LogoIn my opinion, every network analysts should get familiar with these 2 tools, especially Wireshark. In this new version, Wireshark team has fixed few security vulnerabilities, and there is an experimental package for Mac OSX Intel as well.

Other than these releases, I have done a small upgrade from FreeBSD 6.2R to FreeBSD 6.3R recently, which I refer to here. This is only applicable if you are running GENERIC FreeBSD kernel.

[[email protected] /]# fetch

[[email protected] /]# fetch

[[email protected] /]# gpg --verify freebsd-update-upgrade.tgz.asc freebsd-update-upgrade.tgz

[[email protected] /]# tar -xf freebsd-update-upgrade.tgz

[[email protected] /]# sh -f freebsd-update.conf -r 6.3-RELEASE upgrade
Looking up mirrors... 1 mirrors found.
Fetching metadata signature for 6.2-RELEASE from done.
Fetching metadata index... done.
Fetching 1 metadata files... done.
Inspecting system... done.

WARNING: This system is running a "nsm-smpkernel-generic" kernel, which is not a
kernel configuration distributed as part of FreeBSD 6.2-RELEASE.
This kernel will not be updated: you MUST update the kernel manually
before running " install".

The following components of FreeBSD seem to be installed:
src/base src/bin src/contrib src/crypto src/etc src/games src/gnu
src/include src/krb5 src/lib src/libexec src/release src/rescue src/sbin
src/secure src/share src/sys src/tools src/ubin src/usbin world/base
world/catpages world/manpages

The following components of FreeBSD do not seem to be installed:
kernel/generic kernel/smp world/dict world/doc world/games world/info

Does this look reasonable (y/n)? y

Fetching metadata signature for 6.3-RELEASE from done.
Fetching metadata index... done.
Fetching 1 metadata patches. done.
Applying metadata patches... done.
Fetching 1 metadata files... done.
Inspecting system... done.
Preparing to download files... done.
Fetching 6233 patches.....10....20....30.................6230. done.
Applying patches... done.
Fetching 647 files... done.

The following files will be removed as part of updating to 6.3-RELEASE-p1:

[[email protected] /]# sh -f freebsd-update.conf install
[[email protected] /]# shutdown -r now

[[email protected] /]# sh -f freebsd-update.conf install
[[email protected] /]# shutdown -r now

Ourmon to detect UDP Flood

March 15th, 2008

Using graph to detect anomaly has been one of the great features in Ourmon. This is an example why it is great.

If you have been following this blog, you will notice I have 2 posts (including this) that have the strange traffic spike in the nighttime, and it also means that it is good time to attack when everyone is sleeping.
So, there was UDP flood attack on my network in the nighttime yesterday. By checking on Ourmon graphs and reports, I could identify the attack almost immediately without running any other tools.

This is how I identified the UDP flood attack, normally my network will not generate high UDP traffic.

UDP Flood Graph

Do you notice the TCP traffic spike from 2100 to 2200 which is marked in Pink color? I used the same method which I have mentioned previously to check the traffic spike with Argus. It was a inter network file transfer which was running at load 19,001,224 (bits/sec).

Notice the high port ICMP unreachable?

UDP Flood ICMP Unreachable Graph

From UDP Top report, I could see when and which IP launched UDP flood attack, and the time of the attack was stopped.

UDP Top Report

From UDP summarization Top Source IP report:

UDP Top Source IP Report

It is nice that we have not seen any large UDP traffic coming to our network after these IPs were blocked in Firewall.

Ourmon drops packets on 64 bit machine - Fixed

March 10th, 2008

I mentioned I had Ourmon v2.7 installed previously but it generated a lot packets drop at peaks. I thought it was just my machine’s (Quad core Xeon) fault, which is running FreeBSD AMD64 (64 bit).
I have then installed the latest version of Ourmon (v2.8) which has the experimental threaded support and hoped it would run better in threaded but after the compilation and tested for some time, the problem did not seem to go away.

The CPU usage when I run it in T3 (4 threaded), notice the high CPU usage?

48707   root           1    121    0      193M   162M CPU5   5   1:41   93.49% ourmon
48706   root           1    121    0      193M   162M CPU2   2   1:35   91.82% ourmon
48705   root           1    120    0      193M   162M CPU7   7   1:37   90.31% ourmon
48708   root           1    121    0      193M   162M CPU3   3   1:31   90.31% ourmon

Ourmon seemed to generate incorrect packet per second graph for me too:
Packet per second before Fix

The bpfstat information:

  pid   	netif  flags       recv     drop      match      sblen 	   hblen   command
48705     em1  p--s-     240244 67980     240244 16777175 16777098 ourmon
48705     em1  p--s-     270156 86341     270156   248714        0          ourmon

There was same problem when I run Ourmon without threaded support:
i386 usage:

 1100      root        1      -58    0   24932K 24416K bpf    0  37.2H    1.90% ourmon

AMD64 usage:

38236     root        1      109    0     172M   168M  CPU2   2  355:15 70.46% ourmon

I have contacted Jim, the project admin of Ourmon on this issue.
After we have tested for 2 days, I noticed after I have disabled topn_icmperror, topn_scans, and topn_port_scans modules in ourmon.conf, Ourmon has not dropped any packets at peaks.

I reported my findings to Jim and he seemed to find out the problem which caused the Ourmon to drop packet in 64bit machine:

on x86, unsigned int is 4 bytes, unsigned long is 4 bytes
on amd64, unsigned int is 4 bytes, unsigned long is 8 bytes

He sent me a new fixed package and it has been running fine after compilation. Here is the resouce usage with default ourmon.conf on 64 bit:

39548     root        1      -58    0   51164K 29752K    bpf    1  15:29    3.96%  ourmon

The pkts graph is displaying correct result now:
Packet per second After Fixed

Jim said this fix will be included in Ourmon v2.9, and it will come with a couple of new features as well.
While we wait for the new package now, it is time to test the blacklist features in v2.8.
Thanks Jim!

Argus to check traffic spike

March 8th, 2008

When I checked my Ourmon graph today, I noticed same spike happened at the same time in 2 days.
Usually I don’t really pay attention if there is little spike in the nighttime, but it is not normal when it happened continuously at the same time in 2 days.

Late night traffic spike

Since I have full network traffic log, I converted the tcpdump traffic to Argus flow.

 [[email protected] /nsmdir/2008-03-08]# argus -r snort.log.time -w argus.out 

The traffic spike was neither generated from port 25 nor port 80, I used racluster to merge the status records from the same flow without traffic of port 25 and 80, then piped to rasort to sort (-m) the total count of the packet transaction. -s option is to select and show the fields you want to print, -L0 is used to show the labels. Here is the result:

[[email protected] /nsmdir/2008-03-08]# racluster -nnr argus.out -w - - not port 80 and not port 25 | rasort -nnr - -m pkts -s +1ltime +load +bytes +rate -L0 | less

        StartTime           LastTime    	 Flgs   Proto            SrcAddr  Sport    Dir       DstAddr  Dport     TotPkts   TotBytes       State     Load   TotBytes         Rate
   03:27:47.764771    03:33:46.183595  e sD        6     ->   2501219 1031277512   FIN 23018378 1031277512  6978.480957
   03:33:46.708627    03:37:43.800844  e sD        6     ->    526755  543955768   FIN 18354234  543955768  2221.730469
   03:31:51.398870    03:33:15.015036  e sD        6     ->    212332  219148920   FIN 20967134  219148920  2539.365479
   03:21:43.412374    03:41:50.372550  e sD        6    < ? >     68916   56532894   CON 374712.5   56532894    57.098816 is one of the hosts in my network, so from the packet count and total bytes information, I could identify the spike was generated by which host. You may sort based on other criterias to get the result you want, as there might be flows from different IP with little packet count, with large bandwidth transfer rate (load).

To check the traffic, I used ra to show the flow:

[[email protected] /nsmdir/2008-03-08]# ra -nnr argus.out -L0 - host | less

         StartTime    Flgs  Proto            SrcAddr  Sport   Dir            DstAddr  Dport  TotPkts   TotBytes State
   03:27:46.938889  e           6     ->           16       1464   CON
   03:27:47.537244  e           6     ->           22       1964   CON
   03:27:47.764771  e s         6     ->     36705   15095192   CON
   03:27:52.764922  e sD        6     ->     30281   12522592   CON
   03:27:57.765092  e iD        6     ->     14032    5782634   CON
   03:28:02.765262  e sD        6     ->     41270   17023976   CON

This seems like a passive FTP transfer for me (by checking the src ports), so I logged on to server to check for the FTP log to verify:

Mar  8 03:33:10 server ftpd: ([email protected]@ [NOTICE] /home/xxx/public_html/mm//xxx.com_2008.03.08_03-31-02 uploaded  (207309250 bytes, 2444.40KB/sec)
Mar  8 03:33:42 server ftpd: ([email protected]@ [NOTICE] /home/xxx/backup//xxx.com_2008.03.08_03-21-02 uploaded  (891289600 bytes, 2428.22KB/sec)
Mar  8 03:37:39 server ftpd: ([email protected]@ [NOTICE] /home/xxx/backup//xxx.com_2008.03.08_03-21-02.001 uploaded  (514473734 bytes, 2121.01KB/sec)

The log information explains all. There was a new job which was setup recently, running at every 3.30 AM to transfer the backup files to host

I do have traffic grapher to check which ports generated the spike but that is another story which is not worth for a blog post.

Linux or FreeBSD is faster now

March 8th, 2008

I wrote about FreeBSD 7 Release on the other day and mentioned it has 15% better performance than best performing Linux kernel when run on multicore systems. There is a report from Slashdot that mentioned “Linux kernel developer Nick Piggin reran the benchmark today and came to a different conclusion.” Ok, which one is faster now? Do we really care about that?

Sysbench was used to do the benchmark, and mySQL was used for the test.
Enjoy the benchmark result here.

Ourmon 2.8 on FreeBSD 7.0-RELEASE

March 3rd, 2008

FreeBSD LogoFreeBSD Team announced the release of FreeBSD 7.0-REL recently which introduces many new features and improvements. There is significant performance improvement for FreeBSD on those multi core systems and it has 15% better performance than best performing Linux kernel. I can’t wait to turn all my NSM sensors which are running on FreeBSD 6.2-REL to 7.0-REL.

Before that, Ourmon, the network monitoring and anomaly detection system had released version 2.8, which introduces a few exciting features as well. The most wanted THREADED (for me) feature is supported now. Although it is still under experimental and only x86(amd compatible) primitive is used for spinlocks, I will definitely give it a try on Quad core NSM sensor(AMD64) which has Ourmon disabled due to massive packets dropped during peak hours.

Ourmon graph:

The installation is pretty straight forward and easy since you just have to run and change the home net address. Some people may want to change their network interface and web directory during installation, but the rest settings are default or being auto detected.

This is my installation of Ourmon 2.8 on FreeBSD 7.0-REL.
1. Install libpcre from ports. (/usr/ports/devel/pcre)
2. Install libpcap from ports. (/usr/ports/net/libpcap)
3. Install rrdtool from ports (/usr/ports/databases/rrdtool)
4. Install apache from ports. (/usr/ports/www/apache13, 20, or 22)
5. Run ourmon’s

Here is the output.

# ./
configuration script to install ourmon.
note: default is suggested like so: [default]
note: just hit carriage-return for default actions
Would you like to install the ourmon probe? [y]
Front-end configuration phase started ####################
pcap in general needs to be reinstalled from before ourmon install
found pcap lib in /usr/local/lib/libpcap.a - which is a good thing
hit CR to continue:

Would you like to compile/install ourmon? [y]
ourmon build: using make -f Makefile.bsd LIBS=/usr/local/lib/libpcre.a /usr/local/lib/libpcap.a
`ourmon' is up to date.

Next we determine the ourmon config/filter file to use.
By default, we use the local /usr/local/mrourmon/etc/ourmon.conf to provide input filters to ourmon.
WARNING: you should read/edit/understand ourmon.conf!
Do you want to use another ourmon.conf file in some other directory than /usr/local/mrourmon/etc? [n]

Next we suggest one modification to the ourmon.conf file.

If this is a default install, you should change the following config directive:

        topn_syn_homeip network/netmask

and set it to your home network and mask (A.B.C.D/maskbits style)
Do you want to change the topn_syn home network address? [y]
note: the home net address may be a subnet or host address (/32).
enter a home net address and mask. []

Do you want to install the ourmon startup script in the ourmon bin? [y]
WARNING: the default for the interface may not be what you want.
WARNING: use #ifconfig -a to determine interfaces.
Please enter the input interface name to sniff from: [le0]
input interface is le0

Please enter directory for probe output files (mon.lite, etc.): [/usr/local/mrourmon/tmp]
probe output directory name is: /usr/local/mrourmon/tmp

Creating bin/ driver for startup of ourmon. placed in ourmon bin for ourmon front-end/probe startup
./ start

copy the startup script (bin/ to /usr/local/etc/rc.d for boot startup? [y]
ourmon front-end install complete
ourmon front-end build worked

You should now run /usr/local/mrourmon/bin/ to start ourmon

e.g., # /usr/local/mrourmon/bin/ start

You can use stop to stop ourmon

part 2: install the back-end,, etc. (web part)? [y]
Back-end configuration phase started ######################################
We need a local web directory for generated web output.
hint: the webpath given here is a guess: give the CORRECT base web directory with /ourmon at the end
enter absolute web server web path directory: [/usr/local/www/data/ourmon] /usr/local/www/ourmon
your output web path is: /usr/local/www/ourmon

Do you want to create the web directory for ourmon?
HINT: good idea if it doesn't exist. [y]
mkdir: /usr/local/www/ourmon: File exists
ln: web.pages: File exists
cp bard/* /usr/local/www/ourmon/bard
cp /usr/local/mrourmon/bin
cp ombatch*.pl /usr/local/mrourmon/bin
cp /usr/local/mrourmon/bin
cp /usr/local/mrourmon/bin
chmod +x /usr/local/mrourmon/bin/*.sh
chmod +x /usr/local/mrourmon/bin/*.pl

INFO only: also setting up logging directory (if needed)
creating log rrddata tmp dirs, if necessary, in /usr/local/mrourmon
hit CR to continue:

If different, enter front-end output file directory absolute path: [/usr/local/mrourmon/tmp]
probe output file path (back-end input/s) is /usr/local/mrourmon/tmp

Now we copy supplied .html files to the web directory for later editing
do you want to copy base web files to the web directory? [y]

INFO only: setting up local rrdbase directory at /usr/local/mrourmon/rrddata
your runtime rrds get stored in this directory, along with the rrd error log file
if you create new BPF filters, check rrdbase/ourmon.log for errors.
hit CR to continue:

We need a UDP weight threshold for UDP scan alerts
what should the weight be (default is given): [10000000]

Install backend crontab commands in /etc/crontab (default answer y)?: [y]

ourmon system config complete
see INSTALL for post-config sanity checking

This is the error when I tried to start the ourmon.

[[email protected] /usr/local/etc/rc.d]# ./ start
sysctl: unknown oid 'debug.bpf_bufsize'
sysctl: unknown oid 'debug.bpf_maxbufsize'

When I opened the, I found these lines under start_om():

sysctl -w debug.bpf_bufsize=8388608
sysctl -w debug.bpf_maxbufsize=8388608

In FreeBSD, they are called net.bpf.bufsize and net.bpf.maxbufsize. So just change them to these will do.

sysctl -w net.bpf.bufsize=8388608
sysctl -w net.bpf.maxbufsize=8388608

If you have your own value set in /etc/sysctl.conf, just comment these 2 lines.

I will post up as soon as Ourmon with threaded support is successfully setup on AMD64 system.

Say Hello to Network Security

February 29th, 2008

I have been thinking of using what domain name for network security blog, but ended up using this domain.
Why this domain? Because it is my first registered domain name and it means GuTi (that’s me) feels sleepy all the time - zz.

This blog is going to be a place to keep my network security stuffs, and it will allow me to share some thoughts and experiences on system administration, scripting, and forensic related with others.

I am not a good writer, so this is also my best opportunity to improve my English writing skill.

Old story: Threats are everywhere, so having good security practices is very important to protect yourself, your private information, important data, business, website and so on.

Before I end this first post, I would like to say, or print “Hello World!”; (both say and print are available in Perl 6).

Oh ya, I hope I can celebrate birthday for this blog when it turns one year old (4 years later).