This is a live mirror of the Perl 5 development currently hosted at
delete spurious blank lines added by change 23251
[perl5.git] / pod / perlsec.pod
1=head1 NAME
3perlsec - Perl security
425e5e39 7Perl is designed to make it easy to program securely even when running
8with extra privileges, like setuid or setgid programs. Unlike most
54310121 9command line shells, which are based on multiple substitution passes on
425e5e39 10each line of the script, Perl uses a more conventional evaluation scheme
11with fewer hidden snags. Additionally, because the language has more
54310121 12builtin functionality, it can rely less upon external (and possibly
425e5e39 13untrustworthy) programs to accomplish its purposes.
a0d0e21e 14
425e5e39 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
5f05dabc 19mode explicitly by using the B<-T> command line flag. This flag is
425e5e39 20I<strongly> suggested for server programs and any program run on behalf of
fb73857a 21someone else, such as a CGI script. Once taint mode is on, it's on for
22the remainder of your script.
a0d0e21e 23
1e422769 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,
fb73857a 29and it is these checks especially that contribute to making a set-id Perl
425e5e39 30program more secure than the corresponding C program.
fb73857a 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
d929ce6f 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>:
43=over 4
45=item *
47Arguments to C<print> and C<syswrite> are B<not> checked for taintedness.
49=item *
51Symbolic methods
53 $obj->$method(@args);
55and symbolic sub references
57 &{$foo}(@args);
58 $foo->(@args);
60are not checked for taintedness. This requires extra carefulness
61unless you want external data to affect your control flow. Unless
62you carefully limit what these symbolic values are, people are able
63to call functions B<outside> your Perl code, such as POSIX::system,
64in which case they are able to run arbitrary external code.
68For efficiency reasons, Perl takes a conservative view of
69whether data is tainted. If an expression contains tainted data,
70any subexpression may be considered tainted, even if the value
71of the subexpression is not itself affected by the tainted data.
ee556d55 72
d929ce6f 73Because taintedness is associated with each scalar value, some
74elements of an array or hash can be tainted and others not.
75The keys of a hash are never tainted.
a0d0e21e 76
77For example:
425e5e39 79 $arg = shift; # $arg is tainted
80 $hid = $arg, 'bar'; # $hid is also tainted
81 $line = <>; # Tainted
8ebc5c01 82 $line = <STDIN>; # Also tainted
83 open FOO, "/home/me/bar" or die $!;
84 $line = <FOO>; # Still tainted
a0d0e21e 85 $path = $ENV{'PATH'}; # Tainted, but see below
425e5e39 86 $data = 'abc'; # Not tainted
a0d0e21e 87
425e5e39 88 system "echo $arg"; # Insecure
7de90c4d 89 system "/bin/echo", $arg; # Considered insecure
bbd7eb8a 90 # (Perl doesn't know about /bin/echo)
425e5e39 91 system "echo $hid"; # Insecure
92 system "echo $data"; # Insecure until PATH set
a0d0e21e 93
425e5e39 94 $path = $ENV{'PATH'}; # $path now tainted
a0d0e21e 95
54310121 96 $ENV{'PATH'} = '/bin:/usr/bin';
c90c0ff4 97 delete @ENV{'IFS', 'CDPATH', 'ENV', 'BASH_ENV'};
a0d0e21e 98
425e5e39 99 $path = $ENV{'PATH'}; # $path now NOT tainted
100 system "echo $data"; # Is secure now!
a0d0e21e 101
425e5e39 102 open(FOO, "< $arg"); # OK - read-only file
103 open(FOO, "> $arg"); # Not OK - trying to write
a0d0e21e 104
bbd7eb8a 105 open(FOO,"echo $arg|"); # Not OK
425e5e39 106 open(FOO,"-|")
7de90c4d 107 or exec 'echo', $arg; # Also not OK
a0d0e21e 108
425e5e39 109 $shout = `echo $arg`; # Insecure, $shout now tainted
a0d0e21e 110
425e5e39 111 unlink $data, $arg; # Insecure
112 umask $arg; # Insecure
a0d0e21e 113
bbd7eb8a 114 exec "echo $arg"; # Insecure
115 exec "echo", $arg; # Insecure
116 exec "sh", '-c', $arg; # Very insecure!
a0d0e21e 117
118 @files = <*.c>; # insecure (uses readdir() or similar)
119 @files = glob('*.c'); # insecure (uses readdir() or similar)
7bac28a0 120
121 # In Perl releases older than 5.6.0 the <*.c> and glob('*.c') would
122 # have used an external program to do the filename expansion; but in
123 # either case the result is tainted since the list of filenames comes
124 # from outside of the program.
126 $bad = ($arg, 23); # $bad will be tainted
127 $arg, `true`; # Insecure (although it isn't really)
a0d0e21e 129If you try to do something insecure, you will get a fatal error saying
7de90c4d 130something like "Insecure dependency" or "Insecure $ENV{PATH}".
425e5e39 131
132=head2 Laundering and Detecting Tainted Data
134To test whether a variable contains tainted data, and whose use would
135thus trigger an "Insecure dependency" message, you can use the
136tainted() function of the Scalar::Util module, available in your
137nearby CPAN mirror, and included in Perl starting from the release 5.8.0.
595bde10 138Or you may be able to use the following C<is_tainted()> function.
425e5e39 139
140 sub is_tainted {
61890e45 141 return ! eval { eval("#" . substr(join("", @_), 0, 0)); 1 };
425e5e39 142 }
144This function makes use of the fact that the presence of tainted data
145anywhere within an expression renders the entire expression tainted. It
146would be inefficient for every operator to test every argument for
147taintedness. Instead, the slightly more efficient and conservative
148approach is used that if any tainted value has been accessed within the
149same expression, the whole expression is considered tainted.
5f05dabc 151But testing for taintedness gets you only so far. Sometimes you have just
152to clear your data's taintedness. Values may be untainted by using them
153as keys in a hash; otherwise the only way to bypass the tainting
54310121 154mechanism is by referencing subpatterns from a regular expression match.
425e5e39 155Perl presumes that if you reference a substring using $1, $2, etc., that
156you knew what you were doing when you wrote the pattern. That means using
157a bit of thought--don't just blindly untaint anything, or you defeat the
158entire mechanism. It's better to verify that the variable has only good
159characters (for certain values of "good") rather than checking whether it
160has any bad characters. That's because it's far too easy to miss bad
161characters that you never thought of.
425e5e39 162
163Here's a test to make sure that the data contains nothing but "word"
164characters (alphabetics, numerics, and underscores), a hyphen, an at sign,
165or a dot.
54310121 167 if ($data =~ /^([-\@\w.]+)$/) {
425e5e39 168 $data = $1; # $data now untainted
169 } else {
3a2263fe 170 die "Bad data in '$data'"; # log this somewhere
425e5e39 171 }
5f05dabc 173This is fairly secure because C</\w+/> doesn't normally match shell
425e5e39 174metacharacters, nor are dot, dash, or at going to mean something special
175to the shell. Use of C</.+/> would have been insecure in theory because
176it lets everything through, but Perl doesn't check for that. The lesson
177is that when untainting, you must be exceedingly careful with your patterns.
19799a22 178Laundering data using regular expression is the I<only> mechanism for
425e5e39 179untainting dirty data, unless you use the strategy detailed below to fork
180a child of lesser privilege.
182The example does not untaint $data if C<use locale> is in effect,
183because the characters matched by C<\w> are determined by the locale.
184Perl considers that locale definitions are untrustworthy because they
185contain data from outside the program. If you are writing a
186locale-aware program, and want to launder data with a regular expression
187containing C<\w>, put C<no locale> ahead of the expression in the same
188block. See L<perllocale/SECURITY> for further discussion and examples.
190=head2 Switches On the "#!" Line
192When you make a script executable, in order to make it usable as a
193command, the system will pass switches to perl from the script's #!
54310121 194line. Perl checks that any command line switches given to a setuid
3a52c276 195(or setgid) script actually match the ones set on the #! line. Some
54310121 196Unix and Unix-like environments impose a one-switch limit on the #!
3a52c276 197line, so you may need to use something like C<-wU> instead of C<-w -U>
54310121 198under such systems. (This issue should arise only in Unix or
199Unix-like environments that support #! and setuid or setgid scripts.)
3a52c276 200
201=head2 Taint mode and @INC
203When the taint mode (C<-T>) is in effect, the "." directory is removed
204from C<@INC>, and the environment variables C<PERL5LIB> and C<PERLLIB>
205are ignored by Perl. You can still adjust C<@INC> from outside the
206program by using the C<-I> command line option as explained in
207L<perlrun>. The two environment variables are ignored because
208they are obscured, and a user running a program could be unaware that
209they are set, whereas the C<-I> option is clearly visible and
210therefore permitted.
212Another way to modify C<@INC> without modifying the program, is to use
213the C<lib> pragma, e.g.:
215 perl -Mlib=/foo program
217The benefit of using C<-Mlib=/foo> over C<-I/foo>, is that the former
218will automagically remove any duplicated directories, while the later
219will not.
425e5e39 221=head2 Cleaning Up Your Path
1fef88e7 223For "Insecure C<$ENV{PATH}>" messages, you need to set C<$ENV{'PATH'}> to a
1e422769 224known value, and each directory in the path must be non-writable by others
225than its owner and group. You may be surprised to get this message even
226if the pathname to your executable is fully qualified. This is I<not>
227generated because you didn't supply a full path to the program; instead,
228it's generated because you never set your PATH environment variable, or
229you didn't set it to something that was safe. Because Perl can't
230guarantee that the executable in question isn't itself going to turn
231around and execute some other program that is dependent on your PATH, it
54310121 232makes sure you set the PATH.
a0d0e21e 233
234The PATH isn't the only environment variable which can cause problems.
235Because some shells may use the variables IFS, CDPATH, ENV, and
236BASH_ENV, Perl checks that those are either empty or untainted when
237starting subprocesses. You may wish to add something like this to your
238setid and taint-checking scripts.
240 delete @ENV{qw(IFS CDPATH ENV BASH_ENV)}; # Make %ENV safer
242It's also possible to get into trouble with other operations that don't
243care whether they use tainted values. Make judicious use of the file
244tests in dealing with any user-supplied filenames. When possible, do
fb73857a 245opens and such B<after> properly dropping any special user (or group!)
246privileges. Perl doesn't prevent you from opening tainted filenames for reading,
247so be careful what you print out. The tainting mechanism is intended to
248prevent stupid mistakes, not to remove the need for thought.
425e5e39 250Perl does not call the shell to expand wild cards when you pass B<system>
251and B<exec> explicit parameter lists instead of strings with possible shell
252wildcards in them. Unfortunately, the B<open>, B<glob>, and
54310121 253backtick functions provide no such alternate calling convention, so more
254subterfuge will be required.
425e5e39 255
256Perl provides a reasonably safe way to open a file or pipe from a setuid
257or setgid program: just create a child process with reduced privilege who
258does the dirty work for you. First, fork a child using the special
259B<open> syntax that connects the parent and child by a pipe. Now the
260child resets its ID set and any other per-process attributes, like
261environment variables, umasks, current working directories, back to the
262originals or known safe values. Then the child process, which no longer
263has any special permissions, does the B<open> or other system call.
264Finally, the child passes the data it managed to access back to the
5f05dabc 265parent. Because the file or pipe was opened in the child while running
425e5e39 266under less privilege than the parent, it's not apt to be tricked into
267doing something it shouldn't.
54310121 269Here's a way to do backticks reasonably safely. Notice how the B<exec> is
425e5e39 270not called with a string that the shell could expand. This is by far the
271best way to call something that might be subjected to shell escapes: just
fb73857a 272never call the shell at all.
cb1a09d0 273
a1ce9542 274 use English '-no_match_vars';
275 die "Can't fork: $!" unless defined($pid = open(KID, "-|"));
276 if ($pid) { # parent
277 while (<KID>) {
278 # do something
279 }
280 close KID;
281 } else {
282 my @temp = ($EUID, $EGID);
283 my $orig_uid = $UID;
284 my $orig_gid = $GID;
285 $EUID = $UID;
286 $EGID = $GID;
287 # Drop privileges
288 $UID = $orig_uid;
289 $GID = $orig_gid;
290 # Make sure privs are really gone
291 ($EUID, $EGID) = @temp;
292 die "Can't drop privileges"
293 unless $UID == $EUID && $GID eq $EGID;
294 $ENV{PATH} = "/bin:/usr/bin"; # Minimal PATH.
295 # Consider sanitizing the environment even more.
296 exec 'myprog', 'arg1', 'arg2'
297 or die "can't exec myprog: $!";
298 }
425e5e39 299
fb73857a 300A similar strategy would work for wildcard expansion via C<glob>, although
301you can use C<readdir> instead.
425e5e39 302
303Taint checking is most useful when although you trust yourself not to have
304written a program to give away the farm, you don't necessarily trust those
305who end up using it not to try to trick it into doing something bad. This
fb73857a 306is the kind of security checking that's useful for set-id programs and
425e5e39 307programs launched on someone else's behalf, like CGI programs.
309This is quite different, however, from not even trusting the writer of the
310code not to try to do something evil. That's the kind of trust needed
311when someone hands you a program you've never seen before and says, "Here,
312run this." For that kind of safety, check out the Safe module,
313included standard in the Perl distribution. This module allows the
314programmer to set up special compartments in which all system operations
315are trapped and namespace access is carefully controlled.
317=head2 Security Bugs
319Beyond the obvious problems that stem from giving special privileges to
fb73857a 320systems as flexible as scripts, on many versions of Unix, set-id scripts
425e5e39 321are inherently insecure right from the start. The problem is a race
322condition in the kernel. Between the time the kernel opens the file to
fb73857a 323see which interpreter to run and when the (now-set-id) interpreter turns
425e5e39 324around and reopens the file to interpret it, the file in question may have
325changed, especially if you have symbolic links on your system.
327Fortunately, sometimes this kernel "feature" can be disabled.
328Unfortunately, there are two ways to disable it. The system can simply
fb73857a 329outlaw scripts with any set-id bit set, which doesn't help much.
330Alternately, it can simply ignore the set-id bits on scripts. If the
425e5e39 331latter is true, Perl can emulate the setuid and setgid mechanism when it
332notices the otherwise useless setuid/gid bits on Perl scripts. It does
333this via a special executable called B<suidperl> that is automatically
54310121 334invoked for you if it's needed.
425e5e39 335
fb73857a 336However, if the kernel set-id script feature isn't disabled, Perl will
337complain loudly that your set-id script is insecure. You'll need to
338either disable the kernel set-id script feature, or put a C wrapper around
425e5e39 339the script. A C wrapper is just a compiled program that does nothing
340except call your Perl program. Compiled programs are not subject to the
fb73857a 341kernel bug that plagues set-id scripts. Here's a simple wrapper, written
425e5e39 342in C:
344 #define REAL_PATH "/path/to/script"
54310121 345 main(ac, av)
425e5e39 346 char **av;
347 {
348 execv(REAL_PATH, av);
54310121 349 }
cb1a09d0 350
54310121 351Compile this wrapper into a binary executable and then make I<it> rather
352than your script setuid or setgid.
425e5e39 353
425e5e39 354In recent years, vendors have begun to supply systems free of this
355inherent security bug. On such systems, when the kernel passes the name
fb73857a 356of the set-id script to open to the interpreter, rather than using a
425e5e39 357pathname subject to meddling, it instead passes I</dev/fd/3>. This is a
358special file already opened on the script, so that there can be no race
359condition for evil scripts to exploit. On these systems, Perl should be
360compiled with C<-DSETUID_SCRIPTS_ARE_SECURE_NOW>. The B<Configure>
361program that builds Perl tries to figure this out for itself, so you
362should never have to specify this yourself. Most modern releases of
363SysVr4 and BSD 4.4 use this approach to avoid the kernel race condition.
365Prior to release 5.6.1 of Perl, bugs in the code of B<suidperl> could
366introduce a security hole.
68dc0745 367
368=head2 Protecting Your Programs
370There are a number of ways to hide the source to your Perl programs,
371with varying levels of "security".
373First of all, however, you I<can't> take away read permission, because
374the source code has to be readable in order to be compiled and
375interpreted. (That doesn't mean that a CGI script's source is
376readable by people on the web, though.) So you have to leave the
377permissions at the socially friendly 0755 level. This lets
378people on your local system only see your source.
68dc0745 379
5a964f20 380Some people mistakenly regard this as a security problem. If your program does
68dc0745 381insecure things, and relies on people not knowing how to exploit those
382insecurities, it is not secure. It is often possible for someone to
383determine the insecure things and exploit them without viewing the
384source. Security through obscurity, the name for hiding your bugs
385instead of fixing them, is little security indeed.
387You can try using encryption via source filters (Filter::* from CPAN,
388or Filter::Util::Call and Filter::Simple since Perl 5.8).
389But crackers might be able to decrypt it. You can try using the byte
390code compiler and interpreter described below, but crackers might be
391able to de-compile it. You can try using the native-code compiler
68dc0745 392described below, but crackers might be able to disassemble it. These
393pose varying degrees of difficulty to people wanting to get at your
394code, but none can definitively conceal it (this is true of every
395language, not just Perl).
397If you're concerned about people profiting from your code, then the
398bottom line is that nothing but a restrictive licence will give you
399legal security. License your software and pepper it with threatening
400statements like "This is unpublished proprietary software of XYZ Corp.
401Your access to it does not give you permission to use it blah blah
402blah." You should see a lawyer to be sure your licence's wording will
403stand up in court.
5a964f20 404
405=head2 Unicode
407Unicode is a new and complex technology and one may easily overlook
408certain security pitfalls. See L<perluniintro> for an overview and
409L<perlunicode> for details, and L<perlunicode/"Security Implications
410of Unicode"> for security implications in particular.
412=head2 Algorithmic Complexity Attacks
414Certain internal algorithms used in the implementation of Perl can
415be attacked by choosing the input carefully to consume large amounts
416of either time or space or both. This can lead into the so-called
417I<Denial of Service> (DoS) attacks.
419=over 4
421=item *
423Hash Function - the algorithm used to "order" hash elements has been
424changed several times during the development of Perl, mainly to be
425reasonably fast. In Perl 5.8.1 also the security aspect was taken
426into account.
428In Perls before 5.8.1 one could rather easily generate data that as
429hash keys would cause Perl to consume large amounts of time because
430internal structure of hashes would badly degenerate. In Perl 5.8.1
431the hash function is randomly perturbed by a pseudorandom seed which
432makes generating such naughty hash keys harder.
433See L<perlrun/PERL_HASH_SEED> for more information.
435The random perturbation is done by default but if one wants for some
436reason emulate the old behaviour one can set the environment variable
437PERL_HASH_SEED to zero (or any other integer). One possible reason
438for wanting to emulate the old behaviour is that in the new behaviour
439consecutive runs of Perl will order hash keys differently, which may
440confuse some applications (like Data::Dumper: the outputs of two
441different runs are no more identical).
504f80c1 442
443B<Perl has never guaranteed any ordering of the hash keys>, and the
444ordering has already changed several times during the lifetime of
445Perl 5. Also, the ordering of hash keys has always been, and
446continues to be, affected by the insertion order.
448Also note that while the order of the hash elements might be
449randomised, this "pseudoordering" should B<not> be used for
450applications like shuffling a list randomly (use List::Util::shuffle()
451for that, see L<List::Util>, a standard core module since Perl 5.8.0;
452or the CPAN module Algorithm::Numerical::Shuffle), or for generating
453permutations (use e.g. the CPAN modules Algorithm::Permute or
454Algorithm::FastPermute), or for any cryptographic applications.
456=item *
458Regular expressions - Perl's regular expression engine is so called
459NFA (Non-Finite Automaton), which among other things means that it can
460rather easily consume large amounts of both time and space if the
461regular expression may match in several ways. Careful crafting of the
462regular expressions can help but quite often there really isn't much
463one can do (the book "Mastering Regular Expressions" is required
464reading, see L<perlfaq2>). Running out of space manifests itself by
465Perl running out of memory.
467=item *
469Sorting - the quicksort algorithm used in Perls before 5.8.0 to
470implement the sort() function is very easy to trick into misbehaving
471so that it consumes a lot of time. Nothing more is required than
472resorting a list already sorted. Starting from Perl 5.8.0 a different
473sorting algorithm, mergesort, is used. Mergesort is insensitive to
474its input data, so it cannot be similarly fooled.
478See L<> for more information,
479and any computer science text book on the algorithmic complexity.
481=head1 SEE ALSO
483L<perlrun> for its description of cleaning up environment variables.