wafamole++

Based off WAF-A-MoLE, a guided mutation-based fuzzer for ML-based Web Application Firewalls (WAFs), inspired by AFL and based on the FuzzingBook by Andreas Zeller et al.

This CLI tool is intended for Machine Learning based WAFs that filter out SQL injections via classifiers. It generates adversarial examples from a base input SQL injection query (provided by the user) that are able to bypass a target WAF.

It can be used to assess and increment the robustness of your WAF - following the instructions for a adapting the custom Model class that wraps around the classifier of your WAF, generating such examples and retraining your classifiers with those.

Dataset training and Model creation

If you'd like to train your own dataset for use in wafamole++, you can use this Google Colab notebook as a reference. This contains all the code used to generate the .dump files used in this tool as models, with the exception of the .json datasets to be used in training.

The original WAF-A-MoLE dataset, available on GitHub was used to train several of the new example models, as well as the SQLiV3.json SQL Injection Dataset from Kaggle.

Mutation operators

All mutation operators are semantics-preserving and use the MySQL implementation of the SQL language.

Below are the mutation operators available in the current version of wafamole++.

Mutation	Example
Case Swapping	admin' OR 1=1# ⇒ admin' oR 1=1#
Whitespace Substitution	admin' OR 1=1# ⇒ admin'\t\rOR\n1=1#
Comment Injection	admin' OR 1=1# ⇒ admin'/**/OR 1=1#
Comment Rewriting	admin'/**/OR 1=1# ⇒ admin'/*xyz*/OR 1=1#abc
Integer Encoding	admin' OR 1=1# ⇒ admin' OR 0x1=(SELECT 1)#
Operator Swapping	admin' OR 1=1# ⇒ admin' OR 1 LIKE 1#
Logical Invariant	admin' OR 1=1# ⇒ admin' OR 1=1 AND 0<1#
Number Shuffling (New!)	admin' OR 1=1# ⇒ admin' OR 2=1#
Base Shuffling (New!)	admin' OR 1=1# ⇒ admin' OR 0x8b=1#
Symbol Injection (New!)	admin' OR 1=1# ⇒ admin'/OR}1=1#

Running WAF-A-MoLE

Prerequisites

• numpy
• keras
• scikit-learn
• joblib
• sqlparse
• networkx
• Click

Setup

• For Debian on WSL 2 (Miniconda Python 3.7)

python setup.py build

python setup.py install

pip install -r requirements.txt

pip install scikit-learn==0.21.1

If this doesn't work, installing cython and a newer version of scikit-learn can fix the issue:

python setup.py build

python setup.py install

pip install -r requirements.txt

pip install cython

pip install scikit-learn==0.21.3

• For Debian on Oracle VM VirtualBox 6.0.24 (Python 3.9)

python3 setup.py build

python3 setup.py install

pip install -r requirements.txt

pip install scikit-learn==0.21.3

Sample Usage

You can evaluate the robustness of your own WAF, or try WAF-A-MoLE against some example classifiers. In the first case, have a look at the Model class. Your custom model needs to implement this class in order to be evaluated by WAF-A-MoLE. We already provide wrappers for sci-kit learn and keras classifiers that can be extend to fit your feature extraction phase (if any).

Help

wafamole --help

Usage: wafamole [OPTIONS] COMMAND [ARGS]...

Options:
  --help  Show this message and exit.

Commands:
  evade  Launch WAF-A-MoLE against a target classifier.

wafamole evade --help

Usage: wafamole evade [OPTIONS] MODEL_PATH PAYLOAD

  Launch WAF-A-MoLE against a target classifier.

Options:
  -T, --model-type TEXT     Type of classifier to load
  -t, --timeout INTEGER     Timeout when evading the model
  -r, --max-rounds INTEGER  Maximum number of fuzzing rounds
  -s, --round-size INTEGER  Fuzzing step size for each round (parallel fuzzing
                            steps)
  --threshold FLOAT         Classification threshold of the target WAF [0.5]
  --random-engine TEXT      Use random transformations instead of evolution
                            engine. Set the number of trials
  --output-path TEXT        Location were to save the results of the random
                            engine. NOT USED WITH REGULAR EVOLUTION ENGINE
  --help                    Show this message and exit.

Evading example models

There are several example models provided, located in wafamole/models/custom/example_models. The classifiers used are listed in the table below.

Classifier name	Algorithm
WafBrain	Recurrent Neural Network
ML-Based-WAF (modified)	Non-Linear Support Vector Machine
Token-based	Naive Bayes
Token-based	Random Forest
Token-based	Linear SVM
Token-based	Gaussian SVM
SQLiGoT - Directed Proportional	Gaussian SVM
SQLiGoT - Directed Unproportional	Gaussian SVM
SQLiGoT - Undirected Proportional	Gaussian SVM
SQLiGoT - Undirected Unproportional	Gaussian SVM

WAF-BRAIN - Recurrent Neural Newtork

Bypass the pre-trained WAF-Brain classifier using a admin' OR 1=1# equivalent.

wafamole evade --model-type waf-brain wafamole/models/custom/example_models/waf-brain.h5  "admin' OR 1=1#"

ML-Based-WAF (modified) - Non-Linear Support Vector Machine

Bypass the pre-trained ML-Based-WAF classifier using a admin' OR 1=1# equivalent.

wafamole evade --model-type svc wafamole/models/custom/svc/svc_trained.dump  "admin' OR 1=1#"

Token-based - Naive Bayes

Bypass the pre-trained token-based Naive Bayes classifier using a admin' OR 1=1# equivalent.

wafamole evade --model-type token wafamole/models/custom/example_models/naive_bayes_trained.dump  "admin' OR 1=1#"

Token-based - Random Forest

Bypass the pre-trained token-based Random Forest classifier using a admin' OR 1=1# equivalent.

wafamole evade --model-type token wafamole/models/custom/example_models/random_forest_trained.dump  "admin' OR 1=1#"

Token-based - Linear SVM

Bypass the pre-trained token-based Linear SVM classifier using a admin' OR 1=1# equivalent.

wafamole evade --model-type token wafamole/models/custom/example_models/lin_svm_trained.dump  "admin' OR 1=1#"

Token-based - Gaussian SVM

Bypass the pre-trained token-based Gaussian SVM classifier using a admin' OR 1=1# equivalent.

wafamole evade --model-type token wafamole/models/custom/example_models/gauss_svm_trained.dump  "admin' OR 1=1#"

Custom adapters

First, create a custom Model class that implements the extract_features and classify methods.

class YourCustomModel(Model):
    def extract_features(self, value: str):
    	# TODO: extract features
        feature_vector = your_custom_feature_function(value)
        return feature_vector

    def classify(self, value):
    	# TODO: compute confidence
        confidence = your_confidence_eval(value)
        return confidence

Then, create an object from the model and instantiate an engine object that uses your model class.

model = YourCustomModel() #your init
engine = EvasionEngine(model)
result = engine.evaluate(payload, max_rounds, round_size, timeout, threshold)

Contribute

As with WAF-A-MoLE, all questions, bug reports and pull requests are welcome.

To further expand upon this project, the following guidelines can be followed:

1. New WAF adapters
2. New mutation operators
3. New search algorithms

Authors

• Henrique Vermelho de Toledo - IC UFRJ Instituto de Computação, Federal University of Rio de Janeiro
• Daigoro Alencar de Oliveira - IC UFRJ Instituto de Computação, Federal University of Rio de Janeiro