This section discusses how to build and install Bro, running it interactively (mostly useful for building up familiarity with the policy language, not for traffic analysis), running it on live and recorded network traffic, modifying Bro policy scripts, and the different run-time flags.
Bro builds on a number of types of Unix: FreeBSD, Solaris, Linux, SunOS, and Digital Unix, though not all versions. It does not build under non-Unix operating systems such as Windows NT.
You can get the latest public release of Bro from the Bro web page, http://www.icir.org/vern/bro.html. Bro is distributed as a gzip'd Unix tar archive, which you can unpack using:
or, on some Unix systems:gzcat
tar-file| tar xf -
tar zxf
tar-file
This creates a subdirectory bro-
XXX-
version,
where XXX is a tag such as pub for public releases and
priv for private releases, and version reflects a version
and possibly a subversion, such as 0.8a20 for version
0.8 and subversion a20.
To build Bro, change to the Bro directory and enter:
./configure
make
Fix me: Need to discuss configuration options here. -enable-brov6
This will compile all of the Bro sources, including a version of the BIND DNS library, version 8, which Bro uses for its non-blocking DNS lookups.
Note: For Linux systems, you may need to use the header files in the linux-include/ subdirectory included in the Bro distribution to successfully compile Bro.
You install Bro by issuing:
make install
Note: I don't actually use this functionality myself, so it does not tend to be well tested and may have bugs.
Bro is written using libpcap the portable packet-capture
library available from
ftp:
[0]//ftp.ee.lbl.gov/
[0]libpcap.tar.Z. While
libpcap knows how to use a wide range of Unix packet filters, it
far and away performs most efficiently used in conjunction with the Berkeley
Packet Filter (BPF) and with BPF descendants (such as the Digital Unix
packet filter). Althought BPF is available from
ftp:
[0]//ftp.ee.lbl.gov/
[0]bpf.tar.Z, installing
it involves modifying your kernel, and perhaps requires significant porting
work. However, it comes as part of several operating systems, such as
FreeBSD, NetBSD, and OpenBSD.
For BPF systems, you should be aware of the follwoing tuning and configuration issues:
Under FreeBSD, the configuration variable to increase is debug.bpf_bufsize, which you can set via sysctl. We recommend creating a script run at boot-up time that increases it from its small default value to something on the order of 100 KB-2 MB, depending on how fast (heavily loaded) is the link being monitored, and how much physical memory the monitor machine has at its disposal.
Important note: some versions of libpcapave internal code that limits the maximum buffer size to 32 KB. For these systems, you should apply the patch included in the Bro distribution in the file libpcap.patch .
Finally, once you have increased the buffer sizes, you should check that running Bro does indeed consume the amount of kernel memory you expect. You can do this under FreeBSD using vmstat -m and searching in the output for the summary of BPF memory. You should find that the MemUse statistic goes up by twice the buffer size for every concurrent Bro or tcpdump you run.2.1 The reason the increase is by twice the buffer size is because Bro uses double-buffering to avoid dropping packets when the buffer fills up.
Once you've built Bro, you can run it interactively to try out simple facets of the policy language. Note that in this mode, Bro is not reading network traffic, so you cannot do any traffic analysis; this mode is simply to try out Bro language features.
You run Bro interactively by simply executing ``bro'' without any arguments. It then reads from stdin and writes to stdout.
Try typing the following to it:
print "hello, world";(The end-of-file is critical to remember. It's also a bit annoying for interactive evaluation, but reflects the fact that Bro is not actually meant for interactive use, it simply works as a side-effect of Bro's structure.)
^D
(i.e., end of file)
Bro will respond by printing:
hello, worldto stdout and exiting.
You can further declare variables and print expressions, for example:
global a = telnet; print a, a > ftp; print www.microsoft.com;will print
23/tcp, T 207.46.230.229, 207.46.230.219, 207.46.230.218where 23/tcp reflects the fact that telnet is a predefined variable whose value is TCP port 23, which is larger than TCP port 21 (i.e., ftp); and the DNS name www.microsoft.com currently returns the above three addresses.
You can also define functions:
function top18bits(a: addr): addr { return mask_addr(a, 18); } print top18bits(128.3.211.7);prints
128.3.192.0and even event handlers:
event bro_done() { print "all done!"; }which prints ``all done!'' when Bro exits.
Usually, rather than running Bro interactively you want it to execute a policy script or a set of policy scripts. You do so by specifying the names of the scripts as command-line arguments, such as:
bro ~/my-policy.bro ~/my-additional-policy.broBro provides several mechanisms for simplifying how you specify which policies to run.
First, if a policy file doesn't exist then it will try again using .bro as a suffix, so the above could be specified as:
bro ~/my-policy ~/my-additional-policy
Second, Bro consults the colon-separated search path $BROPATH to locate policy scripts. If your home directory was listed in $BROPATH, then you could have invoked it above using:
bro my-policy my-additional-policy
Note: If you define $BROPATH, you must include bro-dir/policy, where bro-dir is where you have built or installed Bro, because it has to be able to locate bro-dir/policy/bro.init to initialize itself at run-time.
Third, the @load directive can be used in a policy script to indicate the
Bro should at that point process another policy script (like C's include
directive; see §
@load my-additional-policyand then just invoke Bro using:
bro my-policyproviding you always want to load my-additional-policy whenever you load my-policy.
Note that the predefined Bro module mt loads almost all of the other standard Bro analyzers, so you can pull them in with simply:
@load mtor by invoking Bro using ``bro mt my-policy''.
There are two ways to run Bro on network traffic: on traffic captured live by the network interface(s), and on traffic previously recorded using the -w flag of tcpdump or Bro itself.
Bro reads live traffic from the local network interface whenever you specify the -i flag. As mentioned below, you can specify multiple instances to read from multiple interfaces simultaneously, however the interfaces must all be of the same link type (e.g., you can't mix reading from a Fast Ethernet with reading from an FDDI link, though you can mix a 10 Mbps Ethernet interface with a 100 Mbps Ethernet).
In addition, Bro will read live traffic from the interface(s) listed in the interfaces variable, unless you specify the -r flag (and do not specify -i). So, for example, if your policy script contains:
const interfaces += "sk0"; const interfaces += "sk1";then Bro will read from the sk0 and sk1 interfaces, and you don't need to specify -i.
To run on recorded traffic, you use the -r flag to indicate the trace file Bro should read. As with -i, you can use the flag multiple times to read from multiple files; Bro will merge the packets from the files into a single packet stream based on their timestamps.
The Bro distribution includes an example trace that you can try out, example.ftp-attack.trace. If you invoke Bro using:
setenv BRO_ID example bro -r example.ftp-attack.trace mtyou'll see that it generates a connection summary to stdout, a summary of the FTP sessions to ftp.example, a copy of what would have been real-time alerts had Bro been running on live traffic to log.example, and a summary of unusual traffic anomalies (none in this trace) to weird.example.
One way to alter the policy Bro executes is of course to directly edit the scripts. When this can be avoided, however, that is preferred, and Bro provides a pair of related mechanisms to help you specify refinements to existing policies in separate files.
The first such mechanism is that you can define multiple handlers for a given event. So, for example, even though the standard ftp analyzer (bro-dir/policy/ftp.bro) defines a handler for ftp_request, you can define another handler if you wish in your own policy script, even if that policy script loads (perhaps indirectly, via the mt module) the ftp analyzer. When the event engine generates an ftp_request event, both handlers will be invoked. Deficiency: Presently, you do not have control over the order in which they are invoked; you also cannot completely override one handler with another, preventing the first from being invoked.
Second, the standard policy scripts are often written in terms of redefinable variables. For example, ftp.bro contains a variable ftp_guest_ids that defines a list of usernames the analyzer will consider to reflect guest accounts:
const ftp_guest_ids = { "anonymous", "ftp", "guest", } &redef;While ``const'' marks this variables as constant at run-time, the attribute ``&redef'' specifies that its value can be redefined.
For example, in your own script you could have:
redef ftp_guest_ids = { "anonymous", "guest", "visitor", "student" };instead. (Note the use of ``redef'' rather than ``const'', to indicate that you realize you are redefining an existing variable.)
In addition, for most types of variables you can specify incremental changes to the variable, either new elements to add or old ones to subtract. For example, you could instead express the above as:
redef ftp_guest_ids += { "visitor", "student" }; redef ftp_guest_ids -= "ftp";The potential advantage of incremental refinements such as these are that if any other changes are made to ftp.bro's original definition, your script will automatically inherit them, rather than replacing them if you used the first definition above (which explicitly lists all four names to include in the variable). Sometimes, however, you don't want this form of inheriting later changes; you need to decide on a case-by-case basis, though our experience is that generally the incremental approach works best.
Finally, the use of prefixes provides a way
to specify a whole set of policy scripts to load in a particular context.
For example, if you set $BRO_PREFIXES to ``dialup'',
then a load of ftp.bro will also load dialup.ftp.bro
automatically (if it exists). See §
When invoking Bro, you can control its behavior using the following flags:
[-h] Generate a help message summarizing Bro's options and environment variables, and exit.
[-i interface]
Add interface to the list of interfaces from which Bro should
read network traffic (§
Note that if no interfaces are specified, then Bro will not read any network traffic. It does not have a notion of a ``default'' interface from which to read.
[-p prefix]
Add prefix to the list of prefixes searched by Bro when @load'ing
a script. You can also, or in addition, use @prefix to specify search
prefixes. See §
[-r readfile]
Add readfile to the list of tcpdump
save files that Bro should read. You can use this flag multiple times to
direct Bro to read from multiple save files; it will merge the packets
read from the different files based on their timestamps. Note that if
the save files contain only packet headers and not contents, then of
course Bro's analysis of them will be limited.
Note that use of -r is mutually exclusive with use of -i. However, you can use -r when running scripts that refine interfaces, in which case the -r option takes precedence and Bro performs off-line analysis.
[-w writefile]
Write a tcpdump save file to the file
writefile. Bro will record all of the packets it captures,
including their contents, except as controlled by calls to
set_record_packets.
Note: One exception is that unless you are analyzing HTTP events (for example, by @load'ing the http analyzer), Bro does not record the contents of HTTP SYN/FIN/RST packets to the trace file. The reason for this is that HTTP FIN packets often contain a large amount of data, which is not of any interest if you are not using HTTP analysis, and due to the very high volume of HTTP traffic at many sites, removing this data can significantly reduce the size of the save file. Deficiency: Clearly, this should not be hardwired into Bro but under user control.
Save files written using -w are of course readable using -r. Accordingly, you will generally want to use -w when running Bro on live network traffic so you can rerun it off-line later to understand any problems that arise, and also to experiment with the effects of changes to the policy scripts.
You can also combine -r with -w to both read a save file(s) and write another. This is of interest when using multiple instances of -r, as it provides a way to merge tcpdump save files.
[-v] Print the version of Bro and exit.
[-F]
Instructs Bro that it must resolve all hostnames out of its
private DNS cache (§
The point behind this option is to ensure that Bro starts quickly, rather than possibly stalling for an indeterminant amount of time resolving a hostname. Fast startup simplifies checkpointing a running Bro--you can start up a new Bro and then killing off the old one shortly after. You'd like this to occur in a manner such that there's no period during which neither Bro is watching the network (the older because you killed it off too early, the newer because it's stuck resolving hostnames).
Deficiency: -F should also force Bro to load all regular expressions
from its regular expression cache (§
[-O] Turns on Bro's optimizer for improving its internal representation of the policy script. Note: Currently, the amount of improvement is modest, and there's (as always) a risk of an optimizer bug introducing errors into the execution of the script, so the optimizer is not enabled by default.
[-P]
Instructs Bro to prime its private DNS cache (§
In addition to resolving DNS lookups, Bro will also compile all of the regular expressions used in the script. Note: Such compilations can take a long time--many minutes--so it is crucial for Bro to cache the results. It does so in the file ``.bro-RE-cache.v1'' in the user's home directory. You can delete this file whenever you want, for example to purge out old entries no longer needed.
Note: You don't actually need to use -P to tell Bro to compile the regular expressions. It does so anyway whenever it encounters one that isn't already in the cache. Likewise, you don't need to use -F to tell Bro to use the regular expression cache--it will do so automatically whenever it can. But it's important that Bro does compilation during -P to help with ensuring that a subsequent invocation of Bro starts up quickly.
Deficiency: Bro should support incrementally compiled regular expressions to avoid the expensive compilation phase altogether.
[-W] Instructs Bro to activate its internal watchdog. The watchdog provides self-monitoring to enable Bro to detect if its processing is wedged.
Bro only activates the watchdog if it is reading live network traffic. The watchdog consists of a periodic timer that fires every WATCHDOG_INTERVAL seconds. (Deficiency: clearly this should be a user-definable value.) At that point, the watchdog checks to see whether Bro is still working on the same packet as it was the last time the watchdog expired. If so, then the watchdog logs this fact along with some information regarding when Bro began processing the current packet and how many events it processed after handling the packet. Finally, it prints the packet drop information for the different interfaces Bro was reading from, and aborts execution.
Bro is also sensitive to the following environment variables:
Default: if you don't set this variable, then Bro uses the path
.:policy:policy/local:/usr/local/lib/broThat is, the current directory, any policy/ and policy/local/ subdirectories, and /usr/local/lib/bro/.
[$BRO_PREFIXES] A colon-separate lists of prefixes that Bro should apply to each name in a @load directive. For a given prefix and load-name, Bro constructs the filename:
prefix.load-name.bro(where it doesn't add .bro if load-name already ends in .bro). It then searches for the filename using $BROPATH and loads it if its found. Furthermore, it repeats this process for all of the other prefixes (left-to-right), and loads each file it finds for the different prefixes. Note: Bro also first attempts to load the filename without any prefix at all. If this load fails, then Bro exits with an error complaining that it can't open the given @load file.
For example, if you set $BRO_PREFIXES to:
mysite:mysite.wanand then issue ``@load ftp'', Bro will attempt to load each of the following scripts in the following order:
ftp.bro mysite.ftp.bro mysite.wan.ftp.bro
Default: if you don't specify a value for $BRO_PREFIXES, it defaults to empty, and for the example above Bro would only attempt to @load ftp.bro.