Internet Info 1994 March

home *** CD-ROM | disk | FTP | other *** search

/ Internet Info 1994 March / Internet Info CD-ROM (Walnut Creek) (March 1994).iso / security / crack_4.1-tar / Docs / readme.txt < prev

Wrap

Text File | 1992-06-25 | 30.3 KB | 853 lines

"Crack Version 4.1" A Sensible Password Checker for Unix Alec D.E. Muffett Unix Software Engineer Aberystwyth, Wales, UK (aem@aber.ac.uk or alec_muffett@hicom.lut.ac.uk) ABSTRACT Crack is a freely available program designed to find standard Unix eight-character DES encrypted passwords by standard guessing tech- niques outlined below. It is written to be flexi- ble, configurable and fast, and to be able to make use of several networked hosts via the Berkeley rsh program (or similar), where possible. 1. Statement of Intent This package is meant as a proving device to aid the con- struction of secure computer systems. Users of Crack are advised that they may get severly hassled by authoritarian type sysadmin dudes if they run Crack without proper author- isation. 2. Introduction to Version 4.0 Crack is now into it's fourth version, and has been reworked extensively to provide extra functionality, and the purpose of this release is to consolidate as much of this new func- tionality into as small a package as possible. To this end, Crack may appear to be less configurable: it has been writ- ten on the assumption that you run a fairly modern Unix, one with BSD functionality, and then patched in order to run on other systems. This, surprisingly enough, has led to neater code, and has made possible the introduction of greater flexibility which supercedes many of the options that could be configured in earlier versions of Crack. In the same vein, some of the older options are now mandatory. These, such as feedback mode and CRACK_PRINTOUT are no longer supported as options and probably never will be again. There is just a lot of wastage in not running with them, and too many dependencies in other functions to bother programming around them. March 3, 1992 - 2 - The user interface is basically identical to the previous versions, although some people have asked about providing X-windows GUI's to Crack, I think it would be a waste of time to do so. Crack has far less options than your ordinary version of /bin/ls. 3. Introduction to Version 4.1 Version 4.1 of the Crack program is an attempt to extend the features introduced in v4.0 and provide hooks for external libraries such as Michael Glad's wonderful UFC crypt() implementation, which (on some platforms) can outperform my fcrypt() by a factor of 3. I have also been burdened with the task of making Crack's memory handling bombproof (hah!) in the vague hope that it will survive running out of memory on small machines.[1] The extensions that I mention above regard the addition of extra primitives to the dictionary processing language which permit the production of more concise dictionaries contain- ing words, more of which are likely to be passwords. The idea is to gain efficiency by removing some of the dross from the generated dictionaries. Crack should (generally) be more disk-space efficient now that the program can spot dictionaries which have been compressed using compress or pack and will uncompress them on the fly as necessary (using zcat or pcat respec- tively).[2] 4. Crack Methodology - Part 1: Internals Crack takes as its input a series of password files and source dictionaries. It merges the dictionaries, turns the password files into a sorted list, and generates lists of possible passwords from the merged dictionary or from infor- mation gleaned about users from the password file. It does _________________________ [1] - or even on large ones. Brian Tompsett at Hull tweaked Crack v3.3 until it could run to completion after filling the swapspace on each of a network of SparcStation2's. Due to restructuring work on v4.0, I have had to write my own sorting algorithm & re- implement all of his tweaks from scratch, and can only hope that I have emulated the bombproofness of this desirable (?) functionality. [2] Note to people who are short on memory or swap: do remember that to do this Crack will have to fork() (via popen()) and might not be able to create the un- compressing process. Hence, if you intend to swaplock your machine, don't compress the dictionaries. Switch this off by editing the Crack shellscript. March 3, 1992 - 3 - not attempt to remedy the problem of allowing users to have guessable passwords, and it should NOT be used in place of getting a really good, secure passwd program replacement.[3] The above paragraphs define the purpose of Crack, and embody a great deal of hard work, screams of Eureka!, drunkeness, and a fair amount of swearing too. There is a lot of think- ing, philosophy, and empirical guesswork behind the way that Crack attacks password files, and although it is not per- fect, I certainly hope that Crack will out-do most of it's competitors. Crack works by making many individual passes over the pass- word entries that you supply to it. Each pass generates password guesses based upon a sequence of rules, supplied to the program by the user. The rules are specified in a simplistic language in the files gecos.rules and dicts.rules, to be found in the Scripts directory. The dis- tinction between these two files will be made clear later. The rules are written as a simple string of characters, with one rule to a line. Blank lines, and comment lines begin- ning with a hash character # are ignored. Trailing whi- tespace is also ignored. The instructions in the rule are followed from left to right, and are applied to the diction- ary words one by one, as the words are loaded. Some simple pattern matching primitives are provided for selection pur- poses, so that if the dictionary word does not match the pattern, it is ignored. This saves on time and memory. Before carrying on, I suggest that you browse through Scripts/dicts.rules, take a look at the rules supplied as defaults, and try to work out what they do. The rules are stored in two different files for two dif- ferent purposes. Rules in Scripts/gecos.rules are applied to data generated by Crack from the pw_gecos and pw_gecos entries of the user's password entry. The data fed to the gecos rules for the user aem, who is Alec David Muffett, Systems would be: aem, Alec, David, Muffett, Systems, and a series of permutations of those words, either re-ordering the words and joining them together (eg: AlecMuffett), or making up new words based on initial letters of one word taken with the rest of another (eg: AMuffett).[4] The entire set of rules in gecos.rules is applied to each of _________________________ [3] See the end of ths document for more information about passwd replacements. [4] - and ASystems and DSystems, and MSystems, etc... because Crack does not differentiate. Hence, care should be taken to check for redundancy when adding new rules, so as not to waste time during the gecos pass. March 3, 1992 - 4 - these words, which creates many more permutations and combi- nations, all of which are tested. Hence testing the pass- word gecos information under Crack v4.0 and upwards takes somewhat longer than previously, but it is far more thorough. After a pass has been made over the data based on gecos information, Crack makes further passes over the password data using successive rules from the Scripts/dicts.rules by loading the whole of Dicts/bigdict file into memory, with the rule being applied to each word from that file. This generates a resident dictionary, which is sorted and uniqued so as to prevent wasting time on repetition. After each pass is completed, the memory used by the resident dictionary is freed up, and (hopefully) re-used when the next dictionary is loaded. The Dicts/bigdict dictionary is created by Crack by merging, sorting, and uniq'ing the source dictionaries, which are to be found in the directory DictSrc and which may also be named in the Crack shellscript, via the $STDDICT variable. (The default value of $STDDICT is /usr/dict/words). The file DictSrc/bad_pws.dat is a dictionary which is meant to provide many of those common but non-dictionary pass- words, such as 12345678 or qwerty. If you wish to provide a dictionary of your own, just copy it into the DictSrc directory (use compress on it if you wish to save space; Crack will unpack it whilst generating the big dictionary) and then delete the contents of the Dicts directory by running Scripts/spotless. Your new dic- tionary will be merged in on the next run. For more informa- tion on dictionary attacks, see the excellent paper called "Foiling the Cracker: A Survey of, and Improvements to, Password Security" by Daniel Klein, available from ftp.sei.cmu.edu in ~/pub/dvk/passwd.*. Also, please read the APPENDIX file supplied with this distribution.[5] Having described the method of cracking, perhaps we should now investigate the algorithm used to overlay the cracking mechanism. _________________________ [5] Extra dictionaries (those detailed in Dan Klein's paper) can be obtained via anonymous FTP from ftp.uu.net (137.39.1.9) as ~/pub/dictionaries.tar.Z; or check an Archie database for other possible sources of dictionaries. March 3, 1992 - 5 - 5. Crack Methodology - Part 2: Feedback Filters As is stated above, Crack permutes and loads dictionaries directly into memory, sorts and uniques them, before attempting to use each of the words as a guess for each users' password. If Crack correctly guesses a password, it marks the user as done and does not waste further time on trying to break that users password. Once Crack has finished a dictionary pass, it sweeps the list of users looking for the passwords it has cracked. It stores the cracked passwords in both plaintext and encrypted forms in a feedback file in the directory Runtime. Feedback files have names of the form Runtime/F*. The purpose of this is so that, when Crack is next invoked, it may recognise passwords that it has successfully cracked before, and filter them from the input to the password cracker. This provides an instant list of crackable users who have not changed their passwords since the last time Crack was run. This list appears in a file with name out* in the $CRACK_OUT directory, or on stdout, if foreground mode is invoked (see Options, below). In a similar vein, when a Crack run terminates normally, it writes out to the feedback file all encrypted passwords that it has NOT succeeded in cracking. Crack will then ignore all of these passwords next time you run it. Obviously, this is not desirable if you frequently change your dictionaries or rules, and so there is a script pro- vided, Scripts/mrgfbk which sorts your feedback files, merges them into one, and optionally removes all traces of 'uncrackable' passwords, so that your next Crack run can have a go at passwords it has not succeeded in breaking before. Mrgfbk is invoked automatically if you run Scripts/spotless. 6. Crack Methodology - Part 3: Execution and Networking Each time Crack is invoked, whether networked or not, it generates a diefile with a name of the form Runtime/D* (for network cracks, this file is generated by RCrack, and is of the form Runtime/DR* which points to a real diefile, named Runtime/RD* - see below for details). These diefiles contain debugging information about the job, and are generated so that all the jobs on the entire network can be called quickly by invoking Scripts/plaster. Diefiles delete themselves after they have been run. As you will read in the sections below, Crack has a -network option: This is designed to be a simple method of March 3, 1992 - 6 - automatically spreading the load of password cracking out over several machines on a network, preferably if they are connected by some form of networked filestore. When Crack -network is invoked, it filters its input in the ordinary way, and then splits its load up amongst several machines which are specified in the file Scripts/network.conf. This file contains a series of hostnames, power ratings, flags, etc, relevant to the running of Crack on each machine. Crack then calls Scripts/RCrack to use the rsh command (or similar) to invoke Crack on the other hosts. See the RCrack script, and the example network.conf file for details. 7. Installation Crack is one of those most unusual of beasties, a self- installing program. Some people have complained about this apparent weirdness, but it has grown up with Crack ever since the earliest network version, when I could not be bothered to log into several different machines with several different architectures, just in order to build the binaries. Once the necessary configuration options have been set, the executables are created via make by running the Crack shellscript . Crack's configuration lies in two files, the Crack shell script, which contains all the installation specific confi- guration data, and the file Sources/conf.h, which contains configuration options specific to various binary platforms. In the Crack shellscript, you will have to edit the CRACK_HOME variable to the correct value. This variable should be set to an absolute path name (names relative to ~username are OK, so long as you have some sort of csh) through which the directory containing Crack may be accessed on ALL the machines that Crack will be run on. There is a similar variable CRACK_OUT which specifies where Crack should put its output files - by default, this is the same as $CRACK_HOME. You will also have to edit the file Sources/conf.h and work out which switches to enable. Each #define has a small note explaining its purpose. Where I have been in doubt about the portability of certain library functions, usually I have re-written it, so you should be OK. Let me know of your problems, if you have any. If you will be using Crack -network you will then have to generate a Scripts/network.conf file. This contains a list of hostnames to rsh to, what their binary type is (useful when running a network Crack on several different March 3, 1992 - 7 - architectures), a guesstimate of their relative power (take your slowest machine as unary, and measure all others rela- tive to it), and a list of per-host flags to add to those specified on the Crack command line, when calling that host. There is an example of such a file provided in the Scripts directory - take a look at it. If ever you wish to specify a more precise figure as to the relative power of your machines, or you are simply at a loss, play with the command make tests in the source code directory. This can provide you with the number of fcrypt()s that your machine can do per second, which is a number that you can plug into your network.conf as a measure of your machines' power (after rounding the value to an integer). 8. Usage Okay, so, let's assume that you have edited your Crack script, and your Sources/conf.h file, where do you go from here ? Crack [options] [bindir] /etc/passwd [...other passwd files] Crack -network [options] /etc/passwd [...other passwd files] Where bindir is the optional name of the directory where you want the binaries installed. This is useful where you want to be able to run versions of Crack on several different architectures. If bindir does not exist, a warning will be issued, and the directory created. Note: bindir defaults to the name generic if not supplied. Notes for Yellow Pages (NIS) Users: I have occasional queries about how to get Crack running from a YP password file. There are several methods, but by far the simplest is to generate a passwd format file by running:- ypcat passwd > passwd.yp and then running Crack on this file. 9. Options -f Runs Crack in foreground mode, ie: the password cracker is not backgrounded, and messages appear on stdout and stderr as you would expect. This option is only really useful for very small password files, or when you want to put a wrapper script around Crack. March 3, 1992 - 8 - Foreground mode is disabled if you try running Crack -network -f on the command line, because of the insen- sibility of rshing to several machines in turn, waiting for each one to finish before calling the next. How- ever, please read the section about Network Cracking without NFS/RFS, below. -v Sets verbose mode, whereby Crack will print every guess it is trying on a per-user basis. This is a very quick way of flooding your filestore, but useful if you think something is going wrong. -m Sends mail to any user whose password you crack by invoking Scripts/nastygram with their username as an argument. The reason for using the script is so that a degree of flexibility in the format of the mail message is supplied; ie: you don't have to recompile code in order to change the message.[6] -nvalue Sets the process to be nice()ed to value, so, for exam- ple, the switch -n19 sets the Crack process to run at the lowest priority. -network Throws Crack into network mode, in which it reads the Scripts/network.conf file, splits its input into chunks which are sized according to the power of the target machine, and calls rsh to run Crack on that machine. Options for Crack running on the target machine may be supplied on the command line (eg: verbose or recover mode), or in the network.conf file if they pertain to specific hosts (eg: nice() values). -r<pointfile> This is only for use when running in recover mode. When a running Crack starts pass 2, it periodically saves its state in a pointfile, with a name of the form Runtime/P.* This file can be used to recover where you were should a host crash. Simply invoke Crack in exactly the same manner as the last time, with the addition of the -r switch, (eg: -rRuntime/Pfred12345) _________________________ [6] I'm uncertain about the wisdom of mailing someone like this. If someone browses your cracked user's mail somehow, it's like a great big neon sign pointing at the user saying "This Is A Crackable Account - Go For It!". Not to mention the false sense of security it engenders in the System Manager that he's "informed" the user to change his password. What if the user doesn't log on for 3 months? However, so many people have wired it into their own versions of Crack, I sup- pose it must be provided... AEM <1b>9 March 3, 1992 - 9 - switch. Crack will startup and read the file, and jump to roughly where it left off. If you are cracking a very large password file, this can save oodles of time after a crash. If you were running a network Crack, then the jobs will again be spawned onto all the machines of the original Crack. The program will then check that the host it is running on is the same as is mentioned in the pointfile. If it is not, it will quietly die. Thus, assuming that you supply the same input data and do not change your network.conf file, Crack should pick up where it left off. This is a bit inelegant, but it's better than nothing at the moment. The method of error recovery outlined above causes headaches for users who want to do multiprocessing on parallel architectures. Crack is in no way parallel, and because of the way it's structured (reading stdin from shellscript frontends) it is a pain to divide the work amongst several processes via fork()ing. The hack solution to get several copies of Crack run- ning on one machine with n processors at the moment is to insert n copies of the entry for your parallel machine into the Scripts/network.conf file. If you use the -r option in these circumstances however, you will get n copies of the recovered process running, only one of them will have the correct input data. The old solution to this problem (see old documentation if you are interested) has been negated by the intro- duction of feedback mode, so the best bet in this par- ticular situation is to wait until the other jobs are done (and have written out lists of uncrackable pass- words), and then re-start the jobs from scratch. Any- one whose password was not cracked on the first run will be ignored on the second, if they have not changed it since. This is inelegant, but it's the best I can do in the limited time available. 10. Support Scripts The Scripts directory contains a small number of support and utility scripts, some of which are designed to help Crack users check their progress. Briefly, the most useful ones are:- Scripts/shadmrg This is a small (but hopefully readable) script for merging /etc/passwd and /etc/shadow on System V style shadow password systems. It produces the merged data to stdout, and will need redirecting into a file before Crack can work on it. The script is meant to be fairly March 3, 1992 - 10 - lucid, on the grounds that I worry that there are many shadowing schemes out there, and perhaps not all have the same data format. I have not wired this facility into the Crack command itself because the world does NOT revolve around System V yet, regardless of what some people would have me believe, and I believe that the lack of direct support for NIS outlined above, sets a precedent. There are just too many incompatibilities in shadow password schemes for me to hardwire anything. Scripts/plaster which is named after a dumb joke, but is a simple fron- tend to the Runtime/D* diefiles that each copy of the password cracker generates. Invoking Scripts/plaster will kill off all copies of the password cracker you are running, over the network or otherwise. Scripts/status This script rshes to each machine mentioned in the Scripts/network.conf file, and provides some informa- tion about processes and uptime on that machine. This is useful when you want to find out just how well your password crackers are getting on during a Crack -net- work. Scripts/{clean,spotless} These are really just frontends to a makefile. Invoking Scripts/clean tidies up the Crack home directory, and removes probably unwanted files, but leaves the pre- processed dictionary bigdict intact. Scripts/spotless does the same as Scripts/clean but obliterates bigdict and old output files too, and compresses the feedback files into one. Scripts/nastygram This is the shellscript that is invoked by the password cracker to send mail to users who have guessable pass- words, if the -m option is used. Edit it at your lei- sure to suit your system. Scripts/guess2fbk This script takes your out* files as arguments and reformats the 'Guessed' lines into a slightly messy feedback file, suitable for storing with the others. An occasion where this might be useful is when your cracker has guessed many peoples passwords, and then died for some reason (a crash?) before writing out the guesses to a feedback file. Running Scripts/guess2fbk out* >> Runtime/F.new March 3, 1992 - 11 - will save the work that has been done. 11. Network Cracking without NFS/RFS For those users who have some form of rsh command, but do not have a a networked filestore running between hosts, there is now a solution which will allow you to do networked cracking, proposed to me by Brian Tompsett at Hull. Person- ally, I consider the idea to be potty, but it fills in miss- ing functionality in a wonderfully tacky manner. From the documentation above, you will note that Crack will undo the -f (output in foreground) option, if it is invoked with the -network switch at the same time (see the Options section above). This is true, but it does not apply if you specify -f option in the network.conf file. The practical upshot of doing this is that remote copies of Crack can be made to read from stdin and write to stdout over a network link, and thus remote processing is accom- plished. I have tweaked Crack in such a way, therefore, that if the -f option is specified amongst the crack-flags of a host in the network.conf, rather than backgrounding itself on the remote host, the rsh command on the server is backgrounded, and output is written directly to the files on the server's filestore. There are restrictions upon this method, mostly involving the number of processes that a user may run on the server at any one time, and that you will have to collect feedback output together manually (dropping it into the Runtime directory on the server). However, it works. Also, if you try to use rsh as another user, you will suffer problems if rsh insists on reading something from your terminal (eg: a password for the remote account). Also, recovering using checkpointing goes out the window unless you specify the name of the pointfile as it is named on the remote machine. 12. UFC Support and notes on fast crypt() implementations The stdlib version of the crypt() subroutine is incredibly slow. It is a massive bottleneck to the execution of Crack and on typical platforms that you get at universities, it is rare to find a machine which will achieve more than 50 stan- dard crypt() s per second. On low-end diskless worksta- tions, you may expect 2 or 3 per second. It was this slow- ness of the crypt() algorithm which originally supplied much of the security Unix needed.[7] _________________________ [7] See: "Password Security, A Case History" by Bob Morris & Ken Thomson, in the Unix Programmer Docs. March 3, 1992 - 12 - There are now many implementations of faster versions of crypt() to be found on the network. The one supplied with Crack v3.2 and upwards is called fcrypt(). It was origi- nally written in May 1986 by Robert Baldwin at MIT, and is a good version of the crypt() subroutine. I received a copy from Icarus Sparry at Bath University, who had made a couple of portability enhancements to the code. I rewrote most of the tables and the KeySchedule generating algorithm in the original fdes-init.c to knock 40% off the execution overhead of fcrypt() in the form that it was shipped to me. I inlined a bunch of stuff, put it into a single file, got some advice from Matt Bishop and Bob Baldwin [both of whom I am greatly indebted to] about what to do to the xform() routine and to the fcrypt function itself, and tidied up some algorithms. I have also added more lookup tables and reduced several formula for faster use. Fcrypt() is now barely recognisable as being based on its former incarnation, and it is 3x faster. On a DecStation 5000/200, fcrypt() is about 16 times faster than the standard crypt (your mileage may vary with other architectures and compilers). This speed puts fcrypt() into the "moderately fast" league of crypt implementations. Amongst other crypt implementations available is UFC by Michael Glad. UFC-crypt is a version of the crypt subrou- tine which is optimised for machines with 32-bit long integers and generally outperforms my fcrypt() by a factor of between 1 and 3, for a tradeoff of large memory usage, and memory-cache unfriendliness. Hooks for even more optim- ised assembler versions of crypt() are also provided for some platforms (Sun, HP, ...). Getting UFC to work on 16 bit architectures is nearly impossible. However, on most architectures, UFC generates a stunning increase in the power of Crack, and so, from v4.1 onwards, Crack is written to automatically make use of UFC if it can find it. All that you have to do is to obtain a suitable copy of UFC (preferably a version which mentions that it is compatible with Crack v4.1, and unpack it into a directory called ufc-crypt in $CRACK_HOME, and then delete your old binaries. UFC will then be detected, compiled, tested and used in preference to fcrypt() by the Crack program, wher- ever possible. 13. Conclusions What can be done about brute force attacks on your password file ? You must get a drop-in replacement for the passwd and yppasswd commands; one which will stop people from choosing bad passwords in the first place. There are several March 3, 1992 - 13 - programs to do this; Matt Bishop's passwd+ and Clyde Hoover's npasswd program are good examples which are freely available. Consult an Archie database for more details on where you can get them from. It would be nice if an organisation (such as CERT?) could be persuaded to supply skeletons of sensible passwd commands for the public good, as well as an archive of security related utilities[8] on top of the excellent COPS. However, for Unix security to improve on a global scale, we will also require pressure on the vendors, so that programs are writ- ten correctly from the beginning. _________________________ [8] COPS is available for anonymous FTP from cert.sei.cmu.edu (128.237.253.5) in ~/cops March 3, 1992