I am happy to feature a guest post written by Champ Clark III, the author of Sagan, a real time, “snort like” event and system log sniffing tool. Champ will explain a bit about it and how he made it work with rsyslog. I think it is a very interesting project and I am glad it now has full rsyslog support.
But enough of my words, enjoy the real thing ;)
I admit it, I’m a recent convert to rsyslog. I’ve known about rsyslog for years, but have only recently started using rsyslog in production environments. The primary reason for looking into rsyslog is users of Sagan are requesting support for it. I’m very glad they pushed me in that direction. I knew how popular rsyslog was,
but the ‘hassles’ of changing our core logging facilities seemed like a pain.
I can tell you, it was easy and seamless. Also, after reading Rainer Gerhards’ excellent, “rsyslog: going up from 40K messages per second to 250K“, I knew that I liked this project.
So I bit the bullet, and started working with Sagan and rsyslog. I haven’t looked back since.
I work in the network & computer security field. I’ve known for years the importance of log management. One thing that I had noticed was a lack of open source log & packet levelcorrelation engines. This is essentially what Sagan does. One common comparison of Sagan is Cisco’s MARS. Sagan reads in your logs and attempts to correlate the information with the intrusion detection/prevention system’s packet level information.
At Softwink, Inc, my place of employment, we monitor security events for various clients. At the
packet-level inspection for ‘bad events’ (security related), we use Snort. Snort ‘watches’ the network connectionsand sends out an ‘alert’ when it sees nefarious traffic. We configure Snort to send the ‘alert’ to a MySQL database for further analysis. We can then monitoring these Snort sensors for ‘bad events/attacks’ in real time.
However, we found that we were missing the ‘bigger picture’ without logs. This is where rsyslog and Sagan come into play. Essentially, we take all machines and equipment on a network and forward it to a
centralized server. Rsyslog is the ‘receiver’, and sometimes the sender of these log messages. In many cases, we find that centralized secure logging is a requirement for clients. With rsyslog, we
have the ability to store log information into a MySQL database for archive purposes. We can then give the client access to the log information via Loganalyzer for easy, simple retrieval.
How does Sagan fit into this picture? For security analysis, we only want key, typically security related, events from the logs. Manually searching databases for ‘security related’ events is prone to error. It is easy to ‘miss’ key events. Sagan is the ‘eyes on the logs’ watching for security related events in real time. First, Sagan has to have access to the logs coming into the network. This is very simple with Rsyslog:
# rsyslog.conf file.
# As rsyslog receives logs from remote systems, we put them into a format
# that Sagan can understand:
# We now take the logs, in the above format, and send them to a 'named pipe'
# or FIFO.
Sagan can now ‘read’ the logs as they come into rsyslog from the /var/run/sagan.fifo (named pipe/FIFO) in real time. rsyslog actually performs double duty for us; logging to our MySQLdatabase for archival purposes and handing Sagan log information for analysis.
Over all, there is nothing really new about this concept. However, Sagan does something a bit different than other log analysis engines. When Sagan sees a ‘bad event’, Sagan will log that to your Snort IDS/IPS MySQL/PostgreSQL database. What does this mean? Packet level security events and log events reside in
the same database for correlation. There are several advantages; for one, we can now have a single, unified console for log and IDS/IPS events! Second, we can now take advantage of Snort front-end
software to view log events. For example, if you use BASE or Snorby to view packet level IDS/IPS events, you can use the same software to view log level Sagan events. Maybe your shop uses report generation
tools that query the Snort database to show ‘bad packet events’ in your network. Guess what. You can use those same reporting tools for your log information as well. I’ve posted some example screen shots of Snort & Sagan working together here. The idea is that we take advantage of the Snort community’s work on consoles.
Correlation with Sagan and Snort, at the engine level, works several different ways. First, Sagan can in some cases pull network information directly from the log message and use that for correlation in the SQL database. For example, let’s say an attacker is probing your network and is attempting to get information on the SMTP port. The attacker sends your SMTP server ‘expn root’. Your IDS/IPS engine will ‘detect’ this traffic and
store it. It’ll record the source IP, destination IP, packet dump, time stamp, etc. Sagan will do the same at the log level. Sagan will ‘extract’ as much of the information from the log message for further correlation with the packet level.
Recently, Rainer announced liblognorm (early liblognorm website). This is an exciting project. The idea is to “normalize” log information to a nice, standard usable format. I plan on putting as much support and effort as I can into this project, because it’s an important step. For Sagan, it means we will be able to better
correlate information. In my time to ponder about it since its recent announcement, I can see liblognorm being extremely useful for many different projects.
Sagan also shares another feature with Snort; it uses the same rule sets. Sagan rules sets are very much ‘Snort like’. Here is an example rule (this is a single line, broken just for readability):
alert tcp $EXTERNAL_NET any -> $HOME_NET 22 (msg:"[OPENSSH] Invalid or illegal user";
pcre: "/invalid user|illegal user/i"; classtype: attempted-user;
program: sshd; parse_ip_simple; parse_port_simple; threshold:type limit,
track by_src, count 5, seconds 300; reference:
url,wiki.softwink.com/bin/view/Main/5000022; sid:5000022; rev:4;)
If you’re already a Snort user, then the format and syntax should be very simple to understand. We use ‘pcre’ (regular expressions) to ‘look’ for a message from the program ‘sshd’ that contains the term ‘invalid user’ or ‘illegal user’ (case insensitive). We set the classifications, just as Snort does (for further correlation). We can ‘threshold’ the rule, so we don’t get flooded with events.
Sagan uses this format for a variety of reasons. For one, its a well know format in the security field. Second, we can now take advantage of Snort rule maintenance software! For example ‘oinkmaster’ or ‘pulled pork’. The idea is that with Sagan, you don’t need to ‘re-tool’ your network in order for it to work.
Using Sagan with your Snort based IDS/IPS system is just a feature of Sagan. Sagan can operate independently from Snort databases, and offers the normal bells/whistlers you’d expect in a SEIM (e-mailing alerts, etc).
To tie all this together, it means we can simply monitor packet level threats and log level events from a unified console. We can monitor just about everything in a network from the log level standpoint. We can monitor Cisco gear, Fortigate firewalls, Linux/*nix servers, wireless access points, etc.
Sagan is a relatively new project and still under development. Like rsyslog, Sagan is built from the ground up with performance in mind. Sagan is multi-threaded and written in C with the thought that it should be as efficient with memory and the CPU(s) as possible. Rsyslog seems to follow the same philosophy, yet another reason I made the switch.
The more information you know about a threat to your network/system, the better off you’ll be. That is what the mix of rsyslog and Sagan offers. Throw in IDS/IPS (Snort) monitoring, and you can get a complete view about ‘bad things’ happening in your environment.
For more information about Sagan, please see http://sagan.softwink.com.