Finger in The Power: The Code!

This repository contains code related to the paper The Finger in the Power: How to Fingerprint PCs by Monitoring their Power Consumption.

Prerequisites

Hardware and Software

This fingerprint technique was test only on Intel's and AMD's CPU. If you want to run it on other types of CPUs please check how to measure the power consumption on your machine. Under Linux, you can check your power consumption information by running the command:

• Intel - cat /sys/class/powercap/intel-rapl/intel-rapl:0/intel-rapl:0:0/energy_uj or modprobe msr + rdmsr 0x611 -d
• AMD - modprobe msr + rdmsr 0xc001029b -d

The implementation currently only works on Linux OS.

All the Python code is designed for Python3!

Dependencies

To install the dependencies, please run the following command (adapt it with you packet manager)

For native code:

sudo apt update
sudo apt install python3-pip
sudo apt-get install python3-venv
python3 -m venv env
./env/bin/pip install -r requirements.txt

For portable code:

sudo apt update
sudo apt install firefox, python3-pip, cmake, python3.8
pip install -r requirements.txt
git clone https://github.com/WebAssembly/wabt.git
cd wabt
git submodule update --init
mkdir build
cd build
cmake ..
cmake --build .
export PATH=$PATH:$(pwd) //the path of wabt/build
cd ..
cd .. //the path of web_pow_fp
export PATH=$PATH:$(pwd) //Dont forget to set them in path !
python3 ./wasm_generator.py -l 1000
mkdir build
make

How to collect traces

To run the the power consumption collection , you need to check your architecture first. The architecture should be specified using the three-letter convention. currently, these are the architectures that are supported with the current version of the code: DW CFL CNL CON HSW ICL IVB KBL NHM SKL SKX SNB WOL WSM ZEN+ ZEN2

Then, run the following line:

Native

sudo bash go.sh <CHOSEN ARCHITECTURE> <NUMBER OF TRACES>

example: sudo bash go.sh HSW 1000

Portable

bash go.sh all \\all stands for the entire webAssembly instruction set available for power measuring. 

Classification Notebook

Classification using our method on real data can be viewed using the interactive Python notebook: https://colab.research.google.com/github/FingerInThePower/Finger_In_The_Power/blob/master/finger_in_the_power_classification.ipynb

The raw data is downloaded from: https://drive.google.com/file/d/1RR4VvIQ6jiBneVuPXwdq1E5FK8USI_-U/view?usp=share_link

Most Helpful Instructions

most_helpful_native_instructions.csv and most_helpful_web_assembly_instructions.csv list the 20 most helpful instructions in decreasing order for native and WebAssembly code, respectively, as observed on the on Desk-4590 corpus. Each entry is listed together with a brief summary and the associated instruction set family.

The significance of each instruction used is listed in decreasing order along with its importance score in the raw instructions importance data, which can be found at native_raw_importance_per_instruction.csv for native and web_asm_raw_importance_per_instruction.csv for WebAssembly.

For More Information

Please see the paper's official webpage

Acknowledgement

Some code from this repository, mainly the native component, are based on Gras et al code, available here