Prime+Count: Novel Cross-world Covert Channels on ARM TrustZone

Paper: https://haehyun.github.io/papers/prime+count-acsac18.pdf

Brief

• This repository contains codes for Prime+Count paper.
• You can try to build and test this covert channel between normal world and secure world.
• The source code in this repository is only for the normal world part.
• The secure world part can be implemented by using the same approach depending on your environment.

Prime+Count attack

• Prime+Count implements fast covert channels in single-core and cross-core scenarios in the TrustZone.
• Prime+Count achieves low-noise covert channel via monitoring how many L1/L2 cache refill events happens.
• To get full details, See the paper.

How to build

• See comments in ./build.sh to get usage.
• Run ./build.sh with proper arguments.
• After ./build.sh, See output files under build/
• covert_lkm.ko : kernel module for building covert channel
• main : user application for testing

How to test

• First of all, You have to update secure-os or trusted application running on secure-os.

  $ insmod covert_lkm.ko
  $ ./main 1 1 images/01/data /tmp/01.csv
    (See comments in src/user/main.c to get full usage and example.)
    (images/01/data :  input image to be transfered through Prime+Count covert channel)
    (/tmp/01.csv :  output encoded image, encoded via bucket method)
  

Python script to convert bucket-encoded image to real image

• images/py_scrips/*.py are scripts to convert bucket-encoded image to real image.
• You can try to convert the bucket-encoded image with below instructions.

  $ cd images/py_scripts/test/
  $ python ../collect_image_set_counting_mode.py 01.csv (01.csv is prepared bucket-encoded image for testing)
  $ ls -l output.pbm
    (See output.pbm, and compare it to circle.pbm which is original image.)
  

Image data

• images/ : images for experiments.

Publication

@inproceedings{primecount2018,
  title        = {{Prime+Count: Novel Cross-world Covert Channels on ARM TrustZone}},
  author       = {Cho, Haehyun and Zhang, Penghui and Kim, Donguk and Park, Jinbum and Lee, Choong-hoon and Zhao, Ziming and Doup\'{e}, Adam and Ahn, Gail-Joon},
  booktitle    = {In Proceedings of the 34th Annual Computer Security Applications Conference (ACSAC)},
  month        = Dec,
  year         = 2018,
  address      = {San Juan, Puerto Rico, USA},
}

License

• Under GPLv2.
• Copyrights are specificed under LICENSE.

Contact

• Haehyun Cho haehyun@asu.edu
• Jinbum Park jinb.park@samsung.com, jinb.park7@gmail.com