This is an academic project to detect if a website is vulnerable to e-mail header injection.
If your website has been crawled by our service, and you want to know whether your website is vulnerable or not, do contact us.
###Contact Details:
Dr. Adam Doupe - doupe (at) asu.edu
Sai Prashanth Chandramouli - schand31 (at) asu.edu
This vulnerability is pretty old but has not received much attention. You can read more about it here: [Article about E-Mail header injection] (http://www.damonkohler.com/2008/12/email-injection.html).
Note: Starting point is call_form_parser.py, please check that file, everything from there is marked in the comments.
- CeleryCrawler.py is the config file (kinda), and sets up the queue to execute.
- The call_for_ scripts are basically built to use with Celery, they use the .delay() method to add the tasks asynchronously.
- functions.py is a general purpose library written for this project, but can definitely be reused with other projects.
- form_parser and check_for_email are the backbone scripts of this Crawler, they parse the forms and pull up the forms that have email fields respectively.
- Start the crawler like so, python CeleryCrawler.py worker -l INFO
- And then run the parser, like so python call_form_parser.py
- call_email_form_retriever makes a db query to get the forms with email fields, then reconstructs them into a proper ast structure, and then does a parallel pass to the celery queue.
- fuzzer then constructs the requests with the additional headers.
- The fuzzer has been split into multiple parts now, with malicious and non-malicious payloads.
- This is to save time on fuzzing only the forms that sent us emails in the first place.
- Also, the fuzzer uses a user_name that is set to (reg/mal/n)user followed by the form id.
- Have rerouted all such emails to 3 mailboxes using postfix :D
- Thinking right about now (Jan 9) whether we need the E-Mail analyzer script, seeing as we are only going to nuke them with BCC.
- After discussion with the prof, a minimal email analyzer script that checks for x-dummy-headers, and bcc headers needs to be built.
- The mails in maluser are direct proof of the attack.
- But the mails in normaluser could contain the x-check header, if they do, then that is a successful attack as well.
- This is due to pythons way of attaching headers, instead of overwriting, it ignores duplicate headers, so we need to inject a new one.
- So, for php, (and unduplicated python) we use the malusers file, for the duplicated headers, normaluser is checked.
- The reguser is checked to see which sites actually sent out emails :D, so that we can say, out of X sites, we got Y mails, out of which Z mails were fuzzed
- Have written tests using Unittest module and mocks.
- Checks for all the fuzzer functions, and whether they behave correctly for all inputs.
- Added checks for all form_parser and check_for_email functions.
- NOTE: The tests are in python3, while the code has been down-ported to Python2 to ensure compatibility with the RabbitMQ queue :( So, comment out the Python2 specific lines before running tests --> This is a LOT of work, but Python2 doesnt support the same Module structure as Python3, so rewriting tests is also a pain
Note to self: If you ever write another project in Python3 without EXPLICITLY specifying it, then I will kick you myself, however improbable that may sound. EIBTI !!! remember that, at least in Python :D :D
Note to self: Had to downport to Python2, so this entire thing was rewritten :(