Perl, unlike most other languages, uses arrays that are not declared with a particular length and that may grow and shrink in size as is required by subsequent code. In fact, when assigning a value to an element within the array, if the index provided is beyond the end of the array, the array grows to make it valid. Consider the following example:
my @array = (1, 2, 3); # array initialized print "Array size is $#array\n"; # 2 (index of last element) $array[5] = 0; # array grows so that reference is valid print "Array size is $#array\n"; # 5 my $value = $array[7]; # array unchanged + uninitialized value warning $value = $array[-7]; # array unchanged + uninitialized value warning if (exists $array[9]) { # false, array unchanged print "That's a big array.\n"; } print "Array size is $#array\n"; # still 5 $value = $array[10][0]; # reading a value in list context grows array print "Array size is $#array\n"; # 10!
This automatic growth occurs only if the index provided is positive and the array value is being written, not read, and not passed to a testing function like exists()
or defined()
.
If an attacker is able to substitute a number to be used as an array index and provides the value 1000000000 (1 billion), then Perl will happily try to grow the array to 1 billion elements. Depending on the platform's capabilities, the attempt to grow the array might fail, or hang, or simply cause Perl to consume several gigabytes of memory for the lifetime of the array. Because a consequent denial of service could occur, attackers must not be permitted to control array indices.
Noncompliant Code Example
This noncompliant code example takes a set of users via standard input and adds them to an array, indexed by their UIDs. This program may, for instance, be fed the contents of the /etc/passwd
file.
my @users; while (<STDIN>) { my ($username, $dummy, $uid) = split( /:/); if (not (defined( $uid) and defined( $username))) {next;} if (not $uid =~ /^\d*$/) {next;} $users[$uid] = $username; } # ... Work with @users
This code clearly skips input lines that do not contain a valid UID or user name. It also skips lines where the UID is not a positive number. However, a UID that is large might cause excessive growth of the @users
array and provoke a denial of service.
Compliant Solution
This compliant solution enforces a limit on how large a UID may be. Consequently, the array may not contain more than $max_uid
elements.
my @users; my $max_uid = 10000; while (<STDIN>) { my ($username, $dummy, $uid) = split( /:/); if (not (defined( $uid) and defined( $username))) {next;} if (not $uid =~ /^\d*$/) {next;} if ($uid > $max_uid) {next;} $users[$uid] = $username; } # ... Work with @users
Risk Assessment
Using unsanitized array index values may exhaust memory and cause the program to terminate or hang.
Recommendation | Severity | Likelihood | Remediation Cost | Priority | Level |
---|---|---|---|---|---|
IDS32-PL | low | likely | high | P3 | L3 |
Bibliography
4 Comments
David Svoboda
Yves orton says, via the SEI blog entry:
David Svoboda
Here's a related question: is 'lvalue' an official term in Perl, like it is in C/C++/Java? The Perl documentation does not use it (except for the new feature 'lvalue subroutines'. We try to use official terms in the standard, as that helps maintain clarity and precision of speech. Offhand, I'm not sure that using 'lvalue' informally to try to explain the circumstances in which arrays grow is not immediately helpful,
Brian Hodgins
I think that the following lines of code might be confusing for some:
while
(<STDIN>) {
my
(
$username
,
$dummy
,
$uid
) =
split
( /:/);
split
( /:/). I think this should be considered a little more expansion into something like:while (my $line = <STDIN>) {
my ($username, $dummy, $uid) = split /:/, $line;
...
}
David Svoboda
We considered eschewing usage of
$_
. But it is quite safe to use, in most contexts (see EXP34-PL for one example of insecure usage of$_
).While
chomp()
is a good practice when dissecting a line of text, it is not necessary here, as your typicalpasswd
file will have more than 3 fields.