Section 11.3. Other Integrity Checkers
As powerful and useful as Tripwire Open Source is, it's also complex and CPU-intensive. Furthermore, if you run "commercial" operating systems such as Windows or Solaris, no free version is available. Therefore, two 100% free and open source alternatives to Tripwire are worth mentioning.
The Advanced Intrusion Detection Environment (AIDE) is designed to meet and exceed Tripwire's functionality and is available from http://www.cs.tut.fi/~rammer/aide.html. As of this writing its version number is 0.8, which reflects its youth: this may or may not have performance and stability implications. (For what it's worth, based on recent postings to the AIDE mailing list, AIDE seems to have more compile-time than runtime issues.) AIDE is 100% free to run on any of its supported platforms, whether in commercial or noncommercial settings.
IDS, Forensic Tool, or Both?The premise behind this part of the chapter is that Tripwire and other integrity checkers can act as burglar alarms when run automatically at set intervals. Many people run integrity checkers in this way, including me (admittedly, on a limited scale). But is this a reliable IDS methodology? Not everyone thinks so. In his book Network Intrusion Detection: An Analyst's Handbook, Stephen Northcutt says: "To run a program such as Tripwire once at system build to get a file-integrity baseline is cheap, easy, and smart. To run Tripwire every day is costly because someone has to examine the results of the scan." In other words, in Northcutt's opinion, you shouldn't run Tripwire checks routinely; only after you determine, through other means, that a breach has occurred. This approach limits Tripwire's role to assisting your forensics efforts (i.e., figuring out what happened and which files were affected). I personally think using Tripwire only for forensics makes sense if you have reason to fear attackers skilled enough to trick Tripwire or you have too many servers from which to monitor frequent lengthy Tripwire reports. If either condition applies to you, do further research on the subject and consider a more sophisticated host-based IDS package like the free Linux Intrusion Detection System (LIDS) (http://www.lids.org). Information on LIDS and many other IDS tools can be found in the "Tools" section at http://online.securityfocus.com. |
A less Unix-centric alternative is Fcheck, which is available at http://www.geocities.com/fcheck2000/fcheck.html. Fcheck is a Perl script, which makes it both highly portable and very easy to customize. It's also extremely easy to configure: the configuration file is primarily a list of directories and files to scan and files and subdirectories to exclude. Command-line flags determine which attributes are checked for all of these: Fcheck has an "all or nothing" approach. (For you, that may or may not be a plus.)
On the down side, Fcheck has no built-in cryptographic functionality: unless you configure it to use an external program like md5sum (part of the GNU textutils package), it relies on simple CRC hashes, which are much easier to subvert than cryptographic hashes such as MD5 or Haval. Nor does it encrypt its database as Tripwire does. Fcheck was originally designed with change-control in mind, not security per se.
Accordingly, Fcheck's performance is very fast. While running any integrity checker without cryptographic hash checks is probably a bad idea on high-risk systems, it may be justifiable on systems on which you want a nominal check in place that uses minimal system resources. (Note that Tripwire can be configured this way too.)
Another mitigating factor is frequency of checks: if your integrity checker runs every half hour, an attacker has only 30 minutes to disable or otherwise subvert it before their activity is caught by the checker. Thus, if using noncryptographic hashes makes it feasible for you to run checks more often, this might be a sensible tradeoff. If, on the other hand, the system in question has a large number of local users (i.e., shell accounts), I strongly recommend against it; such users may be able to learn a lot about the system without triggering a violation. The weak hash-check method, insofar as it's ever justifiable, is only good against external attackers.
By the way, running an integrity checker very frequently is not likely to help you catch an attacker "in the act." This is for the simple reason that there is an inevitable lag between the time an integrity checker sends a report and the time when someone actually gets around to reading and responding to it. Rather, the practical value of frequent checks lies in the fact that the more frequently your checker writes reports, the more granularity with which you'll be able to analyze a successful attack after the fact, which may improve your ability to recover from it.
Of the three tools I've covered here, Tripwire is the most mature but also the most encumbered from a software-license perspective. AIDE is completely free, and it has some additional functionality, but is much less mature than Tripwire. Fcheck is fast, free, highly portable, and simple, but also makes some notable tradeoffs at security's expense.
