The World of Computer Software/World_Of_Computer_Software-02-385-Vol-1of3.iso/c/cops_104.zip/cops_104/README.3

Text File | 1992-03-10 | 9KB | 172 lines

Continued Use and Installing COPS
----------------------------------

   Once you are satisfied that COPS indeed does something useful
(hopefully this will occur :-)), a good way to use it is to run it on at
least a semi-regular basis.  Even if it doesn't find any problems
immediately, the problems and holes it detects are of the sort that can
pop up at any given time.  One way of running COPS might be to run it as
an "at" job or by cron (if you run suid.chk via cron, use the "-s" flag
to tell cron where the rest of the COPS programs are, or it will chmod
"/" to mode 700, among other things).

   I strongly advise that whatever directory COPS is placed in be
readable, writable, and executable only by the owner (typing "chmod 700
/usr/foo/bar" or whatever the name is will do this) of the directory.
This is to prevent prying eyes from seeing any security problems your
site may have.  Even if you don't think of them as important, someone
else might come around and change your mind.  Since COPS is fairly
configurable, an intruder could easily change the paths and files that
COPS checks for, hence making it fairly worthless.  Again, this comes
back to the point that COPS is only a tool -- don't put down your
defensive shields merely because COPS says "all clear".  If this sounds
paranoid, it is!  Security people are traditionally paranoid, for a
reason...  In any case, it is probably not a good idea to advertise any
(even) potential weaknesses.

   If you use the shell/C version, typing "make install" will create (if
necessary) a subdirectory with the name you put in $INSTALL_DIR (found
on line 7 of "makefile"); if you run a network with multiple
architectures, you can have several executable versions of COPS in the
same NFS-mounted directory structure.

  You can run COPS with "cops -a archtype", and it will cd into the
archtype directory, use the binaries or config files in that directory
(placed there by a "make install"), and put any results in a
subdirectory of the archtype directory with the appropriate host name.
You can set the secure directory (the directory that COPS finds all of
the programs in and places the results in) by either invoking COPS with
the -s flag (both shell and perl versions), or by setting the $SECURE
variable in the "cops" shell script (line 93.)

   For example, assume you have the following setup, and run COPS with:

machine architecture    hostname    If run COPS with:
=====================   ========    ==================
cray                    ribcage     cops         -s /usr/secure
vax                     bar         cops -a vax  -s /usr/secure
vax                     foo         cops -a vax  -s /usr/secure
sun                     earth       cops -a sun3 -s /usr/secure
sun                     mars        cops -a sun3 -s /usr/secure
sun                     venus       cops -a sun4 -s /usr/secure
mips                    hades       cops         -s /usr/secure

  The resulting directory/reporting structure would be (all reports
would be placed in a file named "year_month_day"):

/usr/secure/cops/ribcage
/usr/secure/cops/vax/bar
/usr/secure/cops/vax/foo
/usr/secure/cops/sun3/earth
/usr/secure/cops/sun3/mars
/usr/secure/cops/sun4/venus
/usr/secure/cops/hades

  Sometimes you will get the same report over and over again, everytime
you run COPS; for instance, with Ultrix 3.x, /dev/kmem is world
readable.  This is a security hole, but many utilities in Ultrix need
this to function.  If you wish to only see reports that are _different_
than the old reports, you first need to have an older report saved in a
file (in $SECURE/hostname, or wherever you usually save the reports).
In the shell version, you can either do:

   cops -m user

or:

   set "MMAIL=YES" (line 55)
   set "ONLY_DIFF=YES" (line 66)

in "cops".  In the perl version, do:

   set "$ONLY_DIFF=1" (line 11 of the config file)

  Every time COPS is run after that, it will compare the report it
generated for the current check with the old report; if it detects any
differences, it will mail you a report.  If not, it will simply discard
it.  This can be a real boon for a site with a lot of machines running
COPS every night.

  Alternately, you can use a "filter_file" to filter out repetative
messages.  There is an example filter file, "cops_filter", that is
included in the package.  It is used by simply typing "cops -f cops_filter",
and can be both very useful and very dangerous.  Useful for obvious
reasons, dangerous because it can cause valid warning messages to be
thrown away before you get to see them.

   There are a couple of further options you may wish to explore.  First
of all, since so many breakins are because of poor password selection by
users, it would be a wise idea to add options to your password checking
program (line 200 in "cops", or line 72 in "cops.cf" for perl users).
You may wish to try some words from a dictionary; you may use either
your system dictionary (usually found in /usr/dict/words), or you may
use the same dictionary that the internet worm found so lucrative when
hitting all those thousands of hosts; that dictionary is in the file
"pass.words".  For example, the way to include the worm dictionary (e.g.
the dictionary r.t.m. used in his Internet Worm) is:

  pass.chk -w pass.words

  Also, try some of the options in the password program, such as "-b",
"-g", "-s", and "-c", which add checks for backward, gecos, single
letter & number, and upper and lower case guesses, respectively.  Of
course, each option will increase the time needed to crack the
passwords, so experiment!  See what is reasonable for your hardware
capabilities and resources.

  I've included the fast crypt functions that are used in Crack; if you
want to try those (highly recommended!), uncomment lines 96-97 in the
makefile, and comment out the normal compile line (95).  Better yet,
get the full Crack package from uunet.uu.net or somewhere else.  It
does a great job at cracking passwords.  I might try to integrate the
whole package at some later time, but it might just be wasted work;
they work well separately.

   By using the "pass_diff.chk" program, you can check only accounts
that have _changed_ their password since the last time you've checked --
this can save enormous amounts of time with large systems.  This way,
you can check your users thoroughly once, then only check them when
their passwords change (possibly to something less secure).  Be careful,
though, if you use this and then later expand your checks and/or the
dictionary you use to search for passwords, since the earlier accounts
that were already checked with an inferior method will not be checked
again until they change their password.  See the file "passwords" in the
"extensions" directory for a replacement "passwd" program that can
disallow poor passwords to begin with.

   The file "is_able.lst" contains a list of files that are to be
checked for world readability and/or writability.  You should look at
this file and add or delete any files you feel are important to your
system.

   After running COPS, if any warnings are given that compromise any
individual user's account (such as a world writable .profile or home
directory, a guessed password, etc.), and the problem is not corrected
immediately (or you are not sure whether or not it is worth hassling the
user to change it), try this:

   If you are using the shell version, edit the file "init_kuang", and
add the compromised user(s) uids and groups in their respective target
lines (below lines 20 and 26, respectively).  If you are running the
perl version, create a file with the compromised users listed in it (see
kuang.1 in the perl distribution).  Now run kuang again to see if the
users can compromise the entire system.  You may change your mind about
not thinking they are a problem!  In addition, kuang does not have to
have "root" as a target (the last line).  Try putting in system
administrators or other powerful figures to see if they are in danger as
well.  If you have "perl" installed on your system, try the perl version
of kuang -- "kuang.pl" (you'll have to unpack the shar file this is
inside -- "kuang.pl.shar", and you may have to edit the first line of the
file "kuang.pl", to reflect where the location that perl is on your system),
because it is a more powerful, faster, and more versitile version.

  That's it!  Congratulations for reading this far :-)  Browse around
the COPS directories for more goodies; the "extra_src", "docs", and
"extensions" directories all have interesting things in them.  Don't
forget to try CARP to analyze your network's data (*ONLY USABLE WITH
"cops -v" RESULT FILES*) , and let me know how it goes.  Finally, good
luck.  Send me ideas, flames, kudos, whatever.

 -- dan