NAME

LWPx::ParanoidAgent - subclass of LWP::UserAgent that protects you from harm

SYNOPSIS

require LWPx::ParanoidAgent;

my $ua = LWPx::ParanoidAgent->new;

# this is 10 seconds overall, from start to finish.  not just between
# socket reads.  and it includes all redirects.  so attackers telling
# you to download from a malicious tarpit webserver can only stall
# you for $n seconds

$ua->timeout(10);

# setup extra block lists, in addition to the always-enforced blocking
# of private IP addresses, loopbacks, and multicast addresses

$ua->blocked_hosts(
   "foo.com",
   qr/\.internal\.company\.com$/i,
   sub { my $host = shift;  return 1 if is_bad($host); },
);

$ua->whitelisted_hosts(
   "brad.lj",
   qr/^192\.168\.64\.3?/,
   sub { ... },
);

# get/set the DNS resolver object that's used
my $resolver = $ua->resolver;
$ua->resolver(Net::DNS::Resolver->new(...));

# and then just like a normal LWP::UserAgent, because it is one.
my $response = $ua->get('http://search.cpan.org/');
...
if ($response->is_success) {
    print $response->content;  # or whatever
}
else {
    die $response->status_line;
}

DESCRIPTION

The LWPx::ParanoidAgent is a class subclassing LWP::UserAgent, but paranoid against attackers. It's to be used when you're fetching a remote resource on behalf of a possibly malicious user.

This class can do whatever LWP::UserAgent can (callbacks, uploads from files, etc), except proxy support is explicitly removed, because in that case you should do your paranoia at your proxy.

Also, the schemes are limited to http and https, which are mapped to LWPx::Protocol::http_paranoid and LWPx::Protocol::https_paranoid, respectively, which are forked versions of the same ones without the "_paranoid". Subclassing them didn't look possible, as they were essentially just one huge function.

This class protects you from connecting to internal IP ranges (unless you whitelist them), hostnames/IPs that you blacklist, remote webserver tarpitting your process (the timeout parameter is changed to be a global timeout over the entire process), and all combinations of redirects and DNS tricks to otherwise tarpit and/or connect to internal resources.

CONSTRUCTOR

new

my $ua = LWPx::ParanoidAgent->new([ %opts ]);

In addition to any constructor options from LWP::UserAgent, you may also set blocked_hosts (to an arrayref), whitelisted_hosts (also an arrayref), and resolver, a Net::DNS::Resolver object.

METHODS

$csr->resolver($net_dns_resolver)
$csr->resolver

Get/set the Net::DNS::Resolver object used to lookup hostnames.

$csr->blocked_hosts(@host_list)
$csr->blocked_hosts

Get/set the list of blocked hosts. The items in @host_list may be compiled regular expressions (with qr//), code blocks, or scalar literals. In any case, the thing that is match, passed in, or compared (respectively), is all of the given hostname, given IP address, and IP address in canonical a.b.c.d decimal notation. So if you want to block "1.2.3.4" and the user entered it in a mix of network/host form in a mix of decimal/octal/hex, you need only block "1.2.3.4" and not worry about the details.

$csr->whitelisted_hosts(@host_list)
$csr->whitelisted_hosts

Like blocked hosts, but matching the hosts/IPs that bypass blocking checks. The only difference is the IP address isn't canonicalized before being whitelisted-matched, mostly because it doesn't make sense for somebody to enter in a good address in a subversive way.

SEE ALSO

See LWP::UserAgent to see how to use this class.

http://contributing.appspot.com/lwpx-paranoidagent http://brad.livejournal.com/2409049.html https://github.com/collectiveintel/LWPx-ParanoidAgent http://search.cpan.org/dist/LWPx-ParanoidAgent

ISSUES

Report issues: https://github.com/collectiveintel/LWPx-ParanoidAgent/issues

WARRANTY

This module is supplied "as-is" and comes with no warranty, expressed or implied. It tries to protect you from harm, but maybe it will. Maybe it will destroy your data and your servers. You'd better audit it and send me bug reports.

BUGS

Maybe. See the warranty above.

COPYRIGHT

Copyright 2005 Brad Fitzpatrick
Copyright 2013 Wes Young (wesyoung.me)

Lot of code from the base class, copyright 1995-2004 Gisle Aas.

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.