Pyhthon PyPi

WebApp -> different broswers, mobile vs desktop

DogGuesser 3000 lol


Software Supply Chain

Fun Unique Stuff
-> Library #1, easy bits, boring bits that go wrong

All these dependencies

Web of dependencies, co-dependencies and inter0dependencies

You dont know who written the code that you are using inside your app
you might not even know the name because of chain of dependencies

What happens when one of these is bad..., do you even know about it?

--------------------------------------------------------------------------------------- phish / recon
SCENARIO: steal secrets.txt using malicious dependency

library that app uses called 'big-lotsa-money'
take over that dependency

Public Python projects have setup.py lisiting
- creator name
- email
- homepage
- URLS 
- links to github repo with keys (private keys) *
- unsecured wikis, confluence pages  *

(scrape Pypi for these links)

Spear-phish these developers to get credentials to PyPi

PyPi has no notifications, no 2FA, no new push versions
----------------------------------------------------------------------------- modification

not actually changing library itself
no malicious code inside app
make our own package, slide it inside complex dependency web
totes_not_evil == 0.0.2

they have no reason to suspect anything suspicious, the end-user is just extra not gonna know
Need to publish as "legacy" sdist

Legit: creating shortcuts or compiling code
Malicious: code execution

Code execution: [while I'm installing go do this other thing !]
Code runs as local user
-------------------------------------------------------------------------------------------- exfil

Trusted PyPi domain, abuse of trust
Create a self-publishing package (package the secrets.txt and push)
bump version, add task, C2 server using the legitimate servers,  if the corp doesnt pay attention....

------------------------------------------------------------------------------------------- IRL example ?
NodeJS Nov 2018 eventstream, millions of millions of downloads everyday



Pyup.io Safety & JFrog X-Ray
- creates dependency map and flags known "bad" packages, they look for libraries that had vulnerability in it
- if you depend on old unpatched version they will tell you can you can update
- doesnt know on new malicious targetted packages
- relies on people to register packages as "bad"

PyCQA Bandit
- looks at code
- parses code into abstract syntax tree
- looks for odd and bad code

- need to download dependencies first
-
-----------------------------------------------------------------------------------------------
what if you're a small team or a student, doesnt know what bad/malicious code looks like
what if you need a one-time quick solution... 


Container installation/code-running bubble


DockEnv 
-> virtual env for Python using Docker
-> safely install and run Python code
-> everything to install and running of code to separate container
-> spits out the text input


Persistent backdoor gets blown away when container finishes
By default the package cant do anything outside what it is executing

Code running in DockEnv only accesses the data you give it
suitable for testing, quick solutions and development

Software Supply Chain, Software Supply Web