If that's correct, then I think you don't quite have the right data structure yet, and your loop is not giving proper treatment to hosts with multiple "Service:..." entries -- e.g. suppose host A has service problems X and Y, host B has just X, and host C has just Y. The way the OP code is written, the output list will have three "vulnerability" units, with a single host in each unit. If the idea is to group all hosts that have a given problem, you missed that goal. Something like the following might work better:

use strict;
use warnings;

$/ = "-------------\n";  # don't bother counting dashes

my %services; # keys will be service messages, values will be arrays o
+f host IPs

while (<>)   # put file name on command line (or pipe data to stdin)
{
    my @service_keys = ();
    my $current_service = '';
    my @lines = split /\n/; # step through line-by-line...
    my $host = '';

    while ( @lines )
    {
        $_ = pop @lines;    # ... from the bottom up
        next if /-------/;
        $current_service = "$_\n$current_service";

        if ( /^Service: / ) {  # top of a sub-record
            push @service_keys, $current_service;
            $current_service = '';
        }
        elsif ( /^Host: +([\d.]+)/ ) { # top of record
            $host = $1;
            push @{$services{$_}}, $host for ( @service_keys );
            last;
        }
    }
    warn "Record $. did not contain a host IP\n"  # (update: die is to
+o harsh here)
        if ( $host eq '' );  # just in case...
}

# %services is now HoA; let's suppose we want
# to order the output by "severity" (high ... low)
# but within any severity level, order doesn't matter:

for my $level ( qw/High Moderate Low/ )  # any others?
{
    for my $service ( grep /Severity: $level/, keys %services )
    {
        print "-----------------\n\nVulnerability:\n\n$service\nHosts:
+\n";
        print join "\n", @{$services{$service}}, "\n";
    }
}
[download]

One thing you might need to be careful about is variable white-space in the input service reports -- e.g. if a given message is reported on various hosts with differences in blank lines or trailing spaces, it will show up as multiple keys in the main hash. But when you start by reading a whole record at a time like you're doing, it's easy to start each iteration by normalizing white-space, if necessary.

(update: when I fudged an extra "Host:" entry to try for two hosts with the same problem, I discovered that the last record in the input file needs to end with "------\n", or else the "split /\n/" needs to be  split /\n/,$_,-1, to preserve trailing newlines, if any; and/or you need to muck with normalizing whitespace anyway.)

Wow! That's exactly what I was trying to do. Thanks to you and the other's who replied. I need to sit down and go through this code line by line and I'm sure I will learn a great deal. I have not learned HoA's yet, so this will be a good time to start.

I know you spent some time on this reply and I am one very grateful monk. I'm going to donate $25 to the monstary in your name.

The line break in your regexp is throwing off your match. If you make that a single line it ends up getting matches. What you end up with is a hash where the keys are "Service ..." all the way to the end of the text block and the values are "Host: <ip>" strings. So your keys are multi-line blocks of text and your values are strings that have IP's. That doesn't seem to be what you'd really want here but maybe it's a start.

You can also put that $/ setting once outside the loop. I find it good practice to local those but it doesn't matter much for a standalone piece of code.

I tested this version:

use strict;
use warnings;

my $file = 'input.txt';

my (%hash, @ips, @alerts);

open (FILE, "$file") or die "Can't open $file\n";
local $/ = '-------------------------------------------------';

while (<FILE>){

   $hash{$2}{$1} = 0 
     if (/(Host: \d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}).*(Service:.*Sever
+ity:.*)/ms);
}

foreach my $key (keys %hash) {
    $hash{$key} = [keys %{$hash{$key}}];
}

my ($key, $values);
while (($key, $values) = each %hash) {
    print "$key --> ", join('|', @$values), "\n";
}
[download]

If you wish to embed whitespace (newlines, spaces, tabs), and '#' initiated comments in a regular expression, you should end the regexp with the /x modifier. See perlre for details.

The /x modifier is very useful because it makes it easy to break regular expressions into visually recognizable chunks, possibly spanning multiple lines.