source: trunk/third/perl/pod/perlsec.pod @ 17035

Revision 17035, 16.6 KB checked in by zacheiss, 23 years ago (diff)
This commit was generated by cvs2svn to compensate for changes in r17034, which included commits to RCS files with non-trunk default branches.
Line 
1=head1 NAME
2
3perlsec - Perl security
4
5=head1 DESCRIPTION
6
7Perl is designed to make it easy to program securely even when running
8with extra privileges, like setuid or setgid programs.  Unlike most
9command line shells, which are based on multiple substitution passes on
10each line of the script, Perl uses a more conventional evaluation scheme
11with fewer hidden snags.  Additionally, because the language has more
12builtin functionality, it can rely less upon external (and possibly
13untrustworthy) programs to accomplish its purposes.
14
15Perl automatically enables a set of special security checks, called I<taint
16mode>, when it detects its program running with differing real and effective
17user or group IDs.  The setuid bit in Unix permissions is mode 04000, the
18setgid bit mode 02000; either or both may be set.  You can also enable taint
19mode explicitly by using the B<-T> command line flag. This flag is
20I<strongly> suggested for server programs and any program run on behalf of
21someone else, such as a CGI script. Once taint mode is on, it's on for
22the remainder of your script.
23
24While in this mode, Perl takes special precautions called I<taint
25checks> to prevent both obvious and subtle traps.  Some of these checks
26are reasonably simple, such as verifying that path directories aren't
27writable by others; careful programmers have always used checks like
28these.  Other checks, however, are best supported by the language itself,
29and it is these checks especially that contribute to making a set-id Perl
30program more secure than the corresponding C program.
31
32You may not use data derived from outside your program to affect
33something else outside your program--at least, not by accident.  All
34command line arguments, environment variables, locale information (see
35L<perllocale>), results of certain system calls (readdir(),
36readlink(), the variable of shmread(), the messages returned by
37msgrcv(), the password, gcos and shell fields returned by the
38getpwxxx() calls), and all file input are marked as "tainted".
39Tainted data may not be used directly or indirectly in any command
40that invokes a sub-shell, nor in any command that modifies files,
41directories, or processes, B<with the following exceptions>:
42
43=over 4
44
45=item *
46
47If you pass a list of arguments to either C<system> or C<exec>,
48the elements of that list are B<not> checked for taintedness.
49
50=item *
51
52Arguments to C<print> and C<syswrite> are B<not> checked for taintedness.
53
54=back
55
56Any variable set to a value
57derived from tainted data will itself be tainted, even if it is
58logically impossible for the tainted data to alter the variable.
59Because taintedness is associated with each scalar value, some
60elements of an array can be tainted and others not.
61
62For example:
63
64    $arg = shift;               # $arg is tainted
65    $hid = $arg, 'bar';         # $hid is also tainted
66    $line = <>;                 # Tainted
67    $line = <STDIN>;            # Also tainted
68    open FOO, "/home/me/bar" or die $!;
69    $line = <FOO>;              # Still tainted
70    $path = $ENV{'PATH'};       # Tainted, but see below
71    $data = 'abc';              # Not tainted
72
73    system "echo $arg";         # Insecure
74    system "/bin/echo", $arg;   # Secure (doesn't use sh)
75    system "echo $hid";         # Insecure
76    system "echo $data";        # Insecure until PATH set
77
78    $path = $ENV{'PATH'};       # $path now tainted
79
80    $ENV{'PATH'} = '/bin:/usr/bin';
81    delete @ENV{'IFS', 'CDPATH', 'ENV', 'BASH_ENV'};
82
83    $path = $ENV{'PATH'};       # $path now NOT tainted
84    system "echo $data";        # Is secure now!
85
86    open(FOO, "< $arg");        # OK - read-only file
87    open(FOO, "> $arg");        # Not OK - trying to write
88
89    open(FOO,"echo $arg|");     # Not OK, but...
90    open(FOO,"-|")
91        or exec 'echo', $arg;   # OK
92
93    $shout = `echo $arg`;       # Insecure, $shout now tainted
94
95    unlink $data, $arg;         # Insecure
96    umask $arg;                 # Insecure
97
98    exec "echo $arg";           # Insecure
99    exec "echo", $arg;          # Secure (doesn't use the shell)
100    exec "sh", '-c', $arg;      # Considered secure, alas!
101
102    @files = <*.c>;             # insecure (uses readdir() or similar)
103    @files = glob('*.c');       # insecure (uses readdir() or similar)
104
105If you try to do something insecure, you will get a fatal error saying
106something like "Insecure dependency" or "Insecure $ENV{PATH}".  Note that you
107can still write an insecure B<system> or B<exec>, but only by explicitly
108doing something like the "considered secure" example above.
109
110=head2 Laundering and Detecting Tainted Data
111
112To test whether a variable contains tainted data, and whose use would thus
113trigger an "Insecure dependency" message, check your nearby CPAN mirror
114for the F<Taint.pm> module, which should become available around November
1151997.  Or you may be able to use the following I<is_tainted()> function.
116
117    sub is_tainted {
118        return ! eval {
119            join('',@_), kill 0;
120            1;
121        };
122    }
123
124This function makes use of the fact that the presence of tainted data
125anywhere within an expression renders the entire expression tainted.  It
126would be inefficient for every operator to test every argument for
127taintedness.  Instead, the slightly more efficient and conservative
128approach is used that if any tainted value has been accessed within the
129same expression, the whole expression is considered tainted.
130
131But testing for taintedness gets you only so far.  Sometimes you have just
132to clear your data's taintedness.  The only way to bypass the tainting
133mechanism is by referencing subpatterns from a regular expression match.
134Perl presumes that if you reference a substring using $1, $2, etc., that
135you knew what you were doing when you wrote the pattern.  That means using
136a bit of thought--don't just blindly untaint anything, or you defeat the
137entire mechanism.  It's better to verify that the variable has only good
138characters (for certain values of "good") rather than checking whether it
139has any bad characters.  That's because it's far too easy to miss bad
140characters that you never thought of.
141
142Here's a test to make sure that the data contains nothing but "word"
143characters (alphabetics, numerics, and underscores), a hyphen, an at sign,
144or a dot.
145
146    if ($data =~ /^([-\@\w.]+)$/) {
147        $data = $1;                     # $data now untainted
148    } else {
149        die "Bad data in $data";        # log this somewhere
150    }
151
152This is fairly secure because C</\w+/> doesn't normally match shell
153metacharacters, nor are dot, dash, or at going to mean something special
154to the shell.  Use of C</.+/> would have been insecure in theory because
155it lets everything through, but Perl doesn't check for that.  The lesson
156is that when untainting, you must be exceedingly careful with your patterns.
157Laundering data using regular expression is the I<only> mechanism for
158untainting dirty data, unless you use the strategy detailed below to fork
159a child of lesser privilege.
160
161The example does not untaint $data if C<use locale> is in effect,
162because the characters matched by C<\w> are determined by the locale.
163Perl considers that locale definitions are untrustworthy because they
164contain data from outside the program.  If you are writing a
165locale-aware program, and want to launder data with a regular expression
166containing C<\w>, put C<no locale> ahead of the expression in the same
167block.  See L<perllocale/SECURITY> for further discussion and examples.
168
169=head2 Switches On the "#!" Line
170
171When you make a script executable, in order to make it usable as a
172command, the system will pass switches to perl from the script's #!
173line.  Perl checks that any command line switches given to a setuid
174(or setgid) script actually match the ones set on the #! line.  Some
175Unix and Unix-like environments impose a one-switch limit on the #!
176line, so you may need to use something like C<-wU> instead of C<-w -U>
177under such systems.  (This issue should arise only in Unix or
178Unix-like environments that support #! and setuid or setgid scripts.)
179
180=head2 Cleaning Up Your Path
181
182For "Insecure C<$ENV{PATH}>" messages, you need to set C<$ENV{'PATH'}> to a
183known value, and each directory in the path must be non-writable by others
184than its owner and group.  You may be surprised to get this message even
185if the pathname to your executable is fully qualified.  This is I<not>
186generated because you didn't supply a full path to the program; instead,
187it's generated because you never set your PATH environment variable, or
188you didn't set it to something that was safe.  Because Perl can't
189guarantee that the executable in question isn't itself going to turn
190around and execute some other program that is dependent on your PATH, it
191makes sure you set the PATH.
192
193The PATH isn't the only environment variable which can cause problems.
194Because some shells may use the variables IFS, CDPATH, ENV, and
195BASH_ENV, Perl checks that those are either empty or untainted when
196starting subprocesses. You may wish to add something like this to your
197setid and taint-checking scripts.
198
199    delete @ENV{qw(IFS CDPATH ENV BASH_ENV)};   # Make %ENV safer
200
201It's also possible to get into trouble with other operations that don't
202care whether they use tainted values.  Make judicious use of the file
203tests in dealing with any user-supplied filenames.  When possible, do
204opens and such B<after> properly dropping any special user (or group!)
205privileges. Perl doesn't prevent you from opening tainted filenames for reading,
206so be careful what you print out.  The tainting mechanism is intended to
207prevent stupid mistakes, not to remove the need for thought.
208
209Perl does not call the shell to expand wild cards when you pass B<system>
210and B<exec> explicit parameter lists instead of strings with possible shell
211wildcards in them.  Unfortunately, the B<open>, B<glob>, and
212backtick functions provide no such alternate calling convention, so more
213subterfuge will be required.
214
215Perl provides a reasonably safe way to open a file or pipe from a setuid
216or setgid program: just create a child process with reduced privilege who
217does the dirty work for you.  First, fork a child using the special
218B<open> syntax that connects the parent and child by a pipe.  Now the
219child resets its ID set and any other per-process attributes, like
220environment variables, umasks, current working directories, back to the
221originals or known safe values.  Then the child process, which no longer
222has any special permissions, does the B<open> or other system call.
223Finally, the child passes the data it managed to access back to the
224parent.  Because the file or pipe was opened in the child while running
225under less privilege than the parent, it's not apt to be tricked into
226doing something it shouldn't.
227
228Here's a way to do backticks reasonably safely.  Notice how the B<exec> is
229not called with a string that the shell could expand.  This is by far the
230best way to call something that might be subjected to shell escapes: just
231never call the shell at all. 
232
233        use English;
234        die "Can't fork: $!" unless defined($pid = open(KID, "-|"));
235        if ($pid) {           # parent
236            while (<KID>) {
237                # do something
238            }
239            close KID;
240        } else {
241            my @temp     = ($EUID, $EGID);
242            my $orig_uid = $UID;
243            my $orig_gid = $GID;
244            $EUID = $UID;
245            $EGID = $GID;
246            # Drop privileges
247            $UID  = $orig_uid;
248            $GID  = $orig_gid;
249            # Make sure privs are really gone
250            ($EUID, $EGID) = @temp;
251            die "Can't drop privileges"
252                unless $UID == $EUID  && $GID eq $EGID;
253            $ENV{PATH} = "/bin:/usr/bin"; # Minimal PATH.
254            # Consider sanitizing the environment even more.
255            exec 'myprog', 'arg1', 'arg2'
256                or die "can't exec myprog: $!";
257        }
258
259A similar strategy would work for wildcard expansion via C<glob>, although
260you can use C<readdir> instead.
261
262Taint checking is most useful when although you trust yourself not to have
263written a program to give away the farm, you don't necessarily trust those
264who end up using it not to try to trick it into doing something bad.  This
265is the kind of security checking that's useful for set-id programs and
266programs launched on someone else's behalf, like CGI programs.
267
268This is quite different, however, from not even trusting the writer of the
269code not to try to do something evil.  That's the kind of trust needed
270when someone hands you a program you've never seen before and says, "Here,
271run this."  For that kind of safety, check out the Safe module,
272included standard in the Perl distribution.  This module allows the
273programmer to set up special compartments in which all system operations
274are trapped and namespace access is carefully controlled.
275
276=head2 Security Bugs
277
278Beyond the obvious problems that stem from giving special privileges to
279systems as flexible as scripts, on many versions of Unix, set-id scripts
280are inherently insecure right from the start.  The problem is a race
281condition in the kernel.  Between the time the kernel opens the file to
282see which interpreter to run and when the (now-set-id) interpreter turns
283around and reopens the file to interpret it, the file in question may have
284changed, especially if you have symbolic links on your system.
285
286Fortunately, sometimes this kernel "feature" can be disabled.
287Unfortunately, there are two ways to disable it.  The system can simply
288outlaw scripts with any set-id bit set, which doesn't help much.
289Alternately, it can simply ignore the set-id bits on scripts.  If the
290latter is true, Perl can emulate the setuid and setgid mechanism when it
291notices the otherwise useless setuid/gid bits on Perl scripts.  It does
292this via a special executable called B<suidperl> that is automatically
293invoked for you if it's needed.
294
295However, if the kernel set-id script feature isn't disabled, Perl will
296complain loudly that your set-id script is insecure.  You'll need to
297either disable the kernel set-id script feature, or put a C wrapper around
298the script.  A C wrapper is just a compiled program that does nothing
299except call your Perl program.   Compiled programs are not subject to the
300kernel bug that plagues set-id scripts.  Here's a simple wrapper, written
301in C:
302
303    #define REAL_PATH "/path/to/script"
304    main(ac, av)
305        char **av;
306    {
307        execv(REAL_PATH, av);
308    }
309
310Compile this wrapper into a binary executable and then make I<it> rather
311than your script setuid or setgid.
312
313In recent years, vendors have begun to supply systems free of this
314inherent security bug.  On such systems, when the kernel passes the name
315of the set-id script to open to the interpreter, rather than using a
316pathname subject to meddling, it instead passes I</dev/fd/3>.  This is a
317special file already opened on the script, so that there can be no race
318condition for evil scripts to exploit.  On these systems, Perl should be
319compiled with C<-DSETUID_SCRIPTS_ARE_SECURE_NOW>.  The B<Configure>
320program that builds Perl tries to figure this out for itself, so you
321should never have to specify this yourself.  Most modern releases of
322SysVr4 and BSD 4.4 use this approach to avoid the kernel race condition.
323
324Prior to release 5.6.1 of Perl, bugs in the code of B<suidperl> could
325introduce a security hole.
326
327=head2 Protecting Your Programs
328
329There are a number of ways to hide the source to your Perl programs,
330with varying levels of "security".
331
332First of all, however, you I<can't> take away read permission, because
333the source code has to be readable in order to be compiled and
334interpreted.  (That doesn't mean that a CGI script's source is
335readable by people on the web, though.)  So you have to leave the
336permissions at the socially friendly 0755 level.  This lets
337people on your local system only see your source.
338
339Some people mistakenly regard this as a security problem.  If your program does
340insecure things, and relies on people not knowing how to exploit those
341insecurities, it is not secure.  It is often possible for someone to
342determine the insecure things and exploit them without viewing the
343source.  Security through obscurity, the name for hiding your bugs
344instead of fixing them, is little security indeed.
345
346You can try using encryption via source filters (Filter::* from CPAN).
347But crackers might be able to decrypt it.  You can try using the
348byte code compiler and interpreter described below, but crackers might
349be able to de-compile it.  You can try using the native-code compiler
350described below, but crackers might be able to disassemble it.  These
351pose varying degrees of difficulty to people wanting to get at your
352code, but none can definitively conceal it (this is true of every
353language, not just Perl).
354
355If you're concerned about people profiting from your code, then the
356bottom line is that nothing but a restrictive licence will give you
357legal security.  License your software and pepper it with threatening
358statements like "This is unpublished proprietary software of XYZ Corp.
359Your access to it does not give you permission to use it blah blah
360blah."  You should see a lawyer to be sure your licence's wording will
361stand up in court.
362
363=head1 SEE ALSO
364
365L<perlrun> for its description of cleaning up environment variables.
Note: See TracBrowser for help on using the repository browser.