Installation
The mail security testing framework works with with Python >=3.5. Just pull this repository and go ahead. No further dependencies are required.

Usage
The script mail-tester.py runs the tests. Read the help message with ./mail-tester.py --help and check the list of test and evasion modules with ./mail-tester.py -l to get an overview about the capabilities and the usage of the script. Some hints:

• At least the parameters --smtp-server and --to should be given for a minimal test run.
• All parameters can also be stored in configuration files without the prefix --. These configuration files can be used by invoking ./mail-tester.py @tester.conf (configuration contained in tester.conf).
• Multiple recipients can be configured with --to for testing of different filter configurations.
• Some mail filtering solutions may reject messages after a while. Use --auto-delay for automatic throttling of the mails. This can be fine-tuned with --delay-step, --delay-max and --delay.
• Some tests (Spam and Malware) require samples. Put these in directories and configure these directories with --spam-folder and --malware-folder parameters. The samples are not included in this repository (and will not be). Good places to get malware are theZoo, Das Malwerk or other collections. Spam can be exported straight from yout Spam folder, but must be in EML format.
• Blacklists can be supplied with the --blacklist parameter and are used as sender addresses.
• The Shellshock and subject XSS test cases should have a valid backconnect domain, where you are able to see any backconnects (especially DNS requests). The free Canary Tokens service can be used for this purpose. Thanks to Thinkst for providing this awesome service!
• Some neat attachment recognition evasion tricks can be enabled with --evasion content-disposition. These were used in the past to confuse AV/sandboxing solutions and let them pass malicious mails.
• Don't forget to log the test results with --log. Mail filtering providers often reject mails in the SMTP dialog, which is reflected in the generated log.
• Test cases can be dumped with --output as plain files in a directory, in MBox (--mbox) or MailDir (--maildir) format. This is useful to test mail user agents without sending any mails, to document or review generated test cases.

Development and Extension

Tests
Own tests can be implemented with a class in one of the iexisting or newly created Python files in the tests/ directory. The class must be a subclass of MailTestBase located in the module tests.base of this project. Newly implemented tests are discovered automatically when the class variable active is set to True. Further (if you plan to contribute tests back to the main repository), the class variables identifier, name and description should be set appropriately.
The following base classes exist with methods or class variables intended for overriding:

• MailTestBase: Test class for generic tests.
  • generateTestCases(): Yields test messages. These should be generated with the MIME* classes from the Python email.mime.* packages or with the Message class from email.message to ensure valid mail messages.
  • active: Boolean value if test should be active.
  • identifier: Short identifier of the test. This one is used to enable or disable tests in parameters.
  • name: Short test title.
  • description: Longer test description, should fit within approximately 100 characters.
  • delivery_sender and delivery_recipient: Boolean values, False by default. Normally, the sender and recipients are set in the message and the Python SMTP module takes them over from there. Sometimes it is desirable to set them explicitely in the SMTP library, which can be configured by setting this values to True.
  • finalizeMessage(msg): By default, the base test class sets the From and To headers accrodingly. This behaviour can be overridden if required for the test case.
• MailAttachmentTestBase: Test class for attachment test cases. This generates a complete valid mail with a Subject and a text part and attaches the test case to it. Derived from MailTestBase, therefore the methods/variables from it can be overridden here, too.
  • generateAttachments(): Yields test cases as (description, attachment) tuples.
  • subject: Sets the subject. The place holder {} is replaced by the description yielded by generateAttachments().
  • generateTestCases(): is already overridden with an implementation of the message generation described above, but may be further adapted if required.

Setting the subjects of generated messages is highly recommended to be able to recongize the tests in the receiving inbox.

Evasions
Evasion classes implement techniques for evading recognition of particular mail properties by mail security solutions. Currently, a evasion technique that tries to hide attachments from such solutions by intentionally broken Content-Disposition headers is implemented.

Implement new Evasions
Evasions are implemented by a factory class pattern. The DeliveryBase class instantiaties a factory class derived from the BaseEvasionFactory class. The factory constructor receives a flag that indicates if the evasion is activated. The evasion factory instance is then passed to the test class and stored in its evasions attribute that contains a dict with the evasion identifiers as keys. Inside the test, a evasion class (based on EvasionBase) is instantiated with getEvasionGenerator(). The constructor parameter are defined individually per evasion technique.
The following base classes are used to implement evasions:

• BaseEvasionFactory: Evasion factories must be based on this class. Usually, only the following class variables should be set:
  • active: Set to True if the evasion should be active.
  • identifier: Short identifier of the evasion module used for enabling it in the test configuration.
  • name: Short title of the evasion technique.
  • description: Longer description of the evasion technique. Should fit in approximately 100 characters.
  • generator_evasion: Evasion class that is instantiated if the evasion is enabled.
  • generator_default: Evasion class that is instantiated if the evasion is disabled.
• BaseEvasion: Implementation of evasions must be a subclass of this base class. The following method must be overridden:
  • __init__(): Should instantiate the class with the base message or attachment that should be manipulated with evasion techniques.
  • generate(): Apply the evasion technique to the object passed to the constructor and yield it to the caller as (description, object with evasion applied) tuple.

Generally, the evasion class should yield all evasion variants and pass the default as dedicated test case, while the default evasion classes only pass the given object or create the required data structures, like headers.

Using Evasion Techniques in Test Cases
Evasion techniques are used in test cases where they are applicable. E.g. if an evasion technique manipulates the header of a mail or attachment, the following steps have to be implemented:

1. Generate the base object (mail or attachment) without consideration of the evasion.
2. Instantiate the appropriate evasion class by utilization of the evasion factory instance from self.evasions, e.g.: evasion_items = self.evasions["evasion_identifier"].getEvasionGenerator(message)
3. Iterate over the generator and yield the test cases:

for evasion_item in evasion_items:
    yield evasion_item

Usage of the Content Disposition Evasion Technique
The content disposition evasion technique is already implemented in the framework and should be used for all test cases that target on the recognition of malicious attachments. The constructor receives an attachment and the intended file name. The evasion class then yields (evasion name, attachment with applied evasion technique) tuples that can directly be yielded by the tests generateAttachments() method.