Vulnerability Analysis

BootStomp – A Bootloader Vulnerability Finder

BootStomp is a boot-loader bug finder. It looks for two different class of bugs: memory corruption and state storage vulnerabilities. For more info please refer to the BootStomp paper at
To run BootStomp’s analyses, please read the following instructions. Note that BootStomp works with boot-loaders compiled for ARM architectures (32 and 64 bits both) and that results might slightly vary depending on angr and Z3’s versions. This is because of the time angr takes to analyze basic blocks and to Z3’s expression concretization results.
Directory structure
  • analysis: Contains analysis results (Ex: IDA idbs etc) of boot images of different devices.
  • tools: Contains tools that can be used to work with various images.
$ pip install angr

How to run it

Run BootStomp using docker
The easiest way to use BootStomp is to run it in a docker container. The folder docker contains an appropriate Dockerfile. These are the commands to use it.

cd docker
# build the docker image
docker build -t bootstomp .
# run the docker image (if you need, use proper options to have persistent changes or shared files)
docker run -it bootstomp

# now you are inside a docker container
cd BootStomp
# run BootStomp's taint analysis on one of the examples
# this will take about 30 minutes
python taint_analysis/ config/config.huawei
# the last line of the output will be something like:
# INFO    | 2017-10-14 01:54:10,617 | _CoreTaint | Results in /tmp/BootloaderTaint_fastboot.img_.out

# you can then "pretty print" the results using:
python taint_analysis/ /tmp/BootloaderTaint_fastboot.img_.out

The output should be something like this:

===================== Start Info path =====================
Dereference address at: 0x5319cL
Reason: at location 0x5319cL a tainted variable is dereferenced and used as address.
Tainted Path 
0x52f3cL -> 0x52f78L -> 0x52f8cL -> 0x52fb8L -> 0x52fc8L -> 0x52fecL -> 0x53000L -> 0x53014L -> 0x5301cL -> 0x53030L -> 0x53044L -> 0x53050L -> 0x5305cL -> 0x53068L
===================== End Info path =====================
# Total sinks related alerts: 5
# Total loop related alerts: 8
# Total dereference related alerts: 4

Run BootStomp manually

Automatic detection of taint sources and sinks

  1. Load the boot-loader binary in IDA (we used v6.95). Depending on the CPU architecture of the phone it has been extracted from, 32 bit or 64 bit IDA is needed.
  2. From the menu-bar, run File => Script file =>
  3. Output will appear in the file taint_source_sink.txt under the same directory as the boot-loader itself.

Configuration file
Create a JSON configuration file for the boot-loader binary (see examples in config/), where:

  • bootloader: boot-loader file path
  • info_path: boot-loader source/sink info file path (i.e., taint_source_sink.txt )
  • arch: architecture’s number of bits (available options are 32 and 64)
  • enable_thumb: consider thumb mode (when needed) during the analysis
  • start_with_thumb: starts the analysis with thumb mode enabled
  • exit_on_dec_error: stop the analysis if some instructions cannot be decoded
  • unlock_addr: unlocking function address. This field is necessary only for finding insecure state storage vulnerabilities.

Finding memory corruption vulnerabilities

python config-file-path

Results will be stored in /tmp/BootloaderTaint_[boot-loader].out, where [boot-loader] is the name of the analyzed boot-loader. Note that paths involving loops might appear more than once.

Finding insecure state storage vulnerability

python config-file-path

Results will be stored in /tmp/UnlockChecker_[boot-loader].out, where [boot-loader] is the name of the analyzed boot-loader. Note that paths involving loops might appear more than once.

Checking results
To check BootStomp results, use the script, as follows:

python results_file

Exploit for CVE-2017-2729

Other references

About the author

Mazen Elzanaty

Add Comment

Click here to post a comment

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: