Differential Fuzzing for Perl Modules
This project aims to demonstrate how we can use the differential fuzzing technique to conduct security analysis in an automated and large-scale way to find security issues in modern components used by applications developed in Perl. Full publication is avaible on: https://heitorgouvea.me/2021/12/08/Differential-Fuzzing-Perl-Libs.
# Download
$ git clone https://github.com/htrgouvea/fuzzpm && cd fuzzpm
# Install libs and dependencies
$ cpanm --installdeps .
Differential fuzzing is an approach where we have our seeds being sent to two or more inputs, where they are consumed and should produce the same output. At the end of the test these outputs are compared, in case of divergence the fuzzer will signal a possible failure [[1]].(https://en.wikipedia.org/wiki/Differential_testing)
So basically we have 3 components:
- Our targets;
- Input seeds;
- Test cases;
Here is a introduction about how you can create your own targets, seeds and test cases.
To create your entire fuzzing case, you first need to create your target library as a package, for example:
package Mojo_URI {
use strict;
use warnings;
use Try::Tiny;
use Mojo::URL;
sub new {
my ($self, $payload) = @_;
try {
my $url = Mojo::URL -> new($payload);
return $url -> host;
}
catch {
return undef;
}
}
}
Store at: ./targets/your-taget-name.pm.
So, you need store your seeds as a file at: ./seeds/your-seeds.txt. And the last part is your case as a YAML file, follow this structure:
test:
seeds:
- path/to/seeds-file.txt
libs:
- First_Target
- Second_Target
- Third_Target
For example, for our first case, the following YAML file was constructed and is supplied to the fuzzer via the parameter “--case”:
test:
seeds:
- seeds/urls-radamsa.txt
libs:
- Mojo_URI
- Tiny_HTTP
- Mojo_UA
- Mechanize
- Lib_Furl
- Simple_URI
$ perl fuzzpm.pl --case cases/json-decode.yml
$ perl fuzzpm.pl --case cases/parsing-url.yml
$ docker build -t fuzzpm .
$ docker run -ti --rm fuzzpm --help
Your contributions and suggestions are heartily ♥ welcome. See here the contribution guidelines. Please, report bugs via issues page and for security issues, see here the security policy. (✿ ◕‿◕)
This work is licensed under MIT License.