nccgroup / android_demystification_toolbox

ADT is a toolset designed to help model application behavior, research and test security vulnerabilities, and facilitate reversing hostile code.

Android Demystification Toolbox

Purpose

ADT is a Miasm-based symbolic execution toolset designed to help model application behavior, research and test security vulnerabilities, and facilitate reversing hostile code.

Features

• The ability to save/restore all threads & processes contexts
• A centralized (json encoded) config file to mimick an Android environment
• A scheduler to handle mutexes, multi-threading and multi-processing
  • fork()
  • pthread_create()
  • pthread_mutex_lock/unlock()
  • Blocking file reads / writes
• Java JNI (java native bridge) implementation which handles
  • Strings
  • Files & directories
  • Arrays
  • Function calls
  • Class fields
• A number of miscallenous implementations
  • sscanf()
  • mktime()
  • crc32, mimicking the c version
  • c++ streams
  • pipes
  • fstat() and android's custom struct w/ padding for aarch64
  • android libc's _system_property_get()
  • varargs (general case scenario working only)
  • prcrl

HOWTO

Setup

• Install Miasm from this Miasm fork. The patches from the experimental branch implement aaarch64 instructions, and provide the corresponding jitter and sandboxing components.
  • git clone https://github.com/nguigo/miasm.git && cd miasm && git checkout experimental
  • pip install .
• git clone https://github.com/nccgroup/android_demystification_toolbox.git
• pip install ./android_demystification_toolbox

General usage

• From the template
  • Fill out the necessary memory requirements in the one_time_setup() function
  • Enter your custom breakpoints in the breakpoints_setup() function
  • Let the main loop take care of handling the multiple contexts
  • Use adt.config to store Android environment information (see samples for more details)
• The NDK sample can be run as a demo with python3 sample.py libhello-jnicallback.so

Saving / Restoring contexts

!! WARNING !!

The contexts are saved and restored using the pickle module, which is not safe with untrusted data. Ensure only trusted contexts are restored.

Usage

• The sample function save_context_and_dump() can be used as a jitter callback:

jitter.add_breakpoint(<my_address>, save_context_and_dump)

• The code automatically looks for a context.pkl to restore, so the following command line can be used:

ln -s context_myaddress.pkl context.pkl

• Upon starting the script, the context will be restored automatically