Link to home
Start Free TrialLog in
Avatar of BrianPap22
BrianPap22

asked on

Securing a Perl login script

I'm not an expert in Perl by any means, but here I went ahead and wrote a tiny login script.  I would like to make the script as secure as possible, both from a malicious user entering in bogus data, and from my own dumb self :)

Standard CGI/Perl stuff here.  Later on I'm going to implement some timing checks against brute force attacks, but I'm not too worried about such a thing, honestly.  I'm also not incredibly concerned with efficiency, but if there is some *major* noob mistake I have made, feel free to point it out and I'll make sure you get some points. Point out as little or as much things as you'd like that is wrong with this script, and I will split the points accordingly.

So without further ado, here it is (I apologize for the poor use of style. I'll get better as I go along). Thank you for your time.


#!/usr/bin/perl

use strict;
use Crypt::PasswdMD5;
use CGI 'param','header';

$CGI::POST_MAX = 128;
$CGI::DISABLE_UPLOADS = 1;

open PWFILE, 'passwords'      # Each line is of "user:MD5" style.
  or die "can't open it";
my @pwfile = <PWFILE>;      # Slurpie it! - not a terribly large file, so no worries.
close PWFILE;

my $user = lc(param("user"));
my $pass = param("pass");

print header( {-type=>'text/plain',-expires=>'now'} );

for (@pwfile) {
     if (
          m/
               ^$user:               # Line begins with "$user:".
               (\$1\$.{8}\$)     # Backreference 1: 12 character salt value: "$1$xxxxxxxx$".
               (.{22})               # Backreference 2: 22 character encrypted password.
               $                    # Line ends.
          /ox
          && $1.$2 eq unix_md5_crypt($pass,$1)     # When we match $user, check the password.
     )
     {
          print "Authorized...";     # Do some quick/short stuff here.
          exit;                              # No more need to continue the search.
     }
}
print "Incorrect username or password."; # If you can read this, then we must not have found a match!
Avatar of BrianPap22
BrianPap22

ASKER

Should I take the lack of responses as a sign that I did something right for once? I find that hard to believe for this, my second Perl script ever.

Would a working prototype help someone? If so, just let me know and I can throw one up on some webspace.

OK, new policy... I won't split the points.  500 points per valid suggestion (I'll just create a new award question). If this is not acceptable under the ToS of EE, then a moderator may kindly let me know, and I will not do that. Otherwise, I have unlimited question points and I'm itchin' to use them :)
Avatar of Tintin
That's a damm good effort for a Perl newbie (I suspect you must have other programming experience).

There's no glaring holes.  Just a couple of comments:

1.  You should add a check to see if $user and/or $pass are not specified, otherwise there's no point in proceeding.  So before the open, do a:

die "No username supplied\n" unless $user;
die "No password supplied\n" unless $pass;

2. Also your for loop could probably be written to be a little more readable (although there's nothing really wrong with it). As I can figure, the password file has entries along the lines of:

username:xxxxxxxxxxxxzzzzzzzzzzzzzzzzzzzzzz

Where x is the salt and z is the encrypted password?  

for (@pwfile) {
  if (/^$user:/) {
      my ($salt,$encpass) = /:(.{12})(.{22})/;
       if ($salt$encpass eq unix_md5_crypt($pass,$salt) {
           print "Authorized\n";
           exit;
       }
  }
}

print "Incorrect username or password\n";    

An example line in the pass file would look like this (bogus of course :)

brian:$1$CuRbwNsF$ibVNqwFpWc9o5fJxwziTFT

$1$CuRbwNsF$ is the salt. The rest is the encoded password. The salt is actually 8 characters, but it always begins with "$1$" and ends with "$".  When you encrypt the plaintext password together with the salt value, it returns the salt value as part of the return (odd that it does it that way, but that's just the way it is).

When I wrote the for() loop, I knew I was going to ask for advice, and I thought the way I did it would actually be the most readable. Of course, that was just my opinion.  But you suspect correctly; I do have other experience. PHP mostly, but I'm a little new to that, also.  I originally wrote this script in PHP, but all of a sudden my needs changed, and I needed it in Perl.

Funny that I forgot to check for empty values. I was trying to keep it as simple as possible for readability.
In that case, change my regex to

my ($salt,$encpass) = /:\$1\$(.{8})\$(.{22})/;
ASKER CERTIFIED SOLUTION
Avatar of ozo
ozo
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
I completely missed that, ozo. I think I'll take the pro-active way and limit their accounts to just letters, numbers, dash and underscore, and that will be that.
> $CGI::POST_MAX = 128;
is more or less useless, unfortunately, but its good to be there anyway ;-)
Reason:  when perl gets called, the web-server already collected all data

As ozo already pointed out: you need to clean all incomming variables ($user, $pass)
Don't know what
  unix_md5_crypt($pass,$1)
does, but if it calls system commands somehow, then I'd try to give as password:
   ';/bin/rm -rf *.*
or:
  |/bin/rm -rf *.*

So please check all parameters using white lists, like:
   $user =~ s#[^a-zA-Z0-9.-]##g;
   $pass =~ s#[^a-zA-Z0-9.-]##g;
take care when adding more characters for the password.
It does not appear to me that unix_md5_crypt() calls system commands.

Read about it here:
http://search.cpan.org/~luismunoz/Crypt-PasswdMD5-1.3/PasswdMD5.pm
"Crypt::PasswdMD5 - Provides interoperable MD5-based crypt() functions"
>  It does not appear to me ..
you know for shure, or not?
Are you shure that underlaying systems (like Digest::MD5) behave this way in future too?
If not, make shure your parameters won't harm.
That's my message, you asked for security .. ;-)
I understand what you're saying.  It probably won't change in the future, but I suppose it won't hurt to prepare for the worst.

I know it's better to use "white lists" instead of "black lists."  So my list of "approved characters" so far is as follows:

a-z
A-Z
0-9
,.<>[]{}?_+
~!@#$%^&*()_+

This should exclude [space], backtick [`], [/\] forward and backslash, pipe [|], colon and semi-colon [;:], and everything else I can't think of.

My user's need to be able to enter in as sophisticated a password as possible, so I don't want to go overkill with this.

You know, I'm starting to get the feeling that I shouldn't be using regular expressions at all. If I set up a for() loop to run through each line and used ($uname,$hash) = split(/:/), then I could just match 'if $uname eq $user'  ... etc.

Might that be safer? Worse?  No difference?  Remember, efficiency is not a priority.
eq $user is more efficient, but otherwise equivalent to =~ m/^\Q$user\E\z/ which are of course safer and more forgiving than m/$user/
with %  you pass over URL encoded charaters
with !  you may access shell commands
with &  you can combine shell commands
with $  you can access variables in shell
with @  you may access functions in M$ SQL
with %_ you have wild cards in SQL
and so on ...

My pedantic filter for a username is:
   a-zA-Z0-9.-
for the password, it needs to be more, I agree
I'll join Tintin on contratulating you for such a nifty first script, and there is only one nit I want to pick:
After your if (regex) { } you have nothing to handle the exception that the format of the password file is invalid (the regex itself is not matched). This is not  actually insecure, as this case would still result in failed authorization, but it is confusing for a user, and hard for an admin to debug, if their correctly typed password should fail due to a corrupted file.
your script is awesome!

MastaLlama