PortPrint

This repository contains a demonstration of the fingerprinting approach described in the paper "PortPrint: PortPrint: Identifying Inaccessible Code with Port Contention" by Hornetz and Schwarz. It records the port contention profile of two AES implementations from Linux and OpenSSL, both using the AES-NI extensions. The final output is a plot of the contention profiles, which should show clear differences between the two implementations.

Prerequisites

To run this demo, you should have a CPU based on Intel's Skylake or Golden Cove architecture, or on AMD's Zen 3 architecture. The demo is meant to run on Linux.

Setup

Building the Recorder

To build the recorder, you will need cmake and libzstd. On Debian-based distributions, you can install them with the following command:

sudo apt-get install cmake libzstd-dev

Then, execute the following command to start building:

cmake -B build -S . && cmake --build build

Setting up the Python Virtual Environment

To avoid conflicts in the Python dependencies, you should set up a virtual environment with the package dependencies. If you have the python3-virtualenv package installed, you can do this with following commands:

python -m virtualenv venv
source venv/bin/activate
pip install -r requirements.txt

Usage

Step 1: Recording Timing Traces

The first step is to record timing information by invoking the recorder with your CPU's architecture. For example, ./recorder Skylake will record timing information with contention primitives for Skylake. Supported values are Skylake, AlderLakeP for CPU's based on Golden Cove, and Zen3. The timing traces will be saved into the measurements folder in you working directory.

Step 2: Pre-processing the traces

Before you can inspect the contention profile, the traces from the previous step must go through a pre-processing step. You can start this by running the preprocess.py script, passing the path to the measurements as an argument. The pre-processed traces will be stored to the processed folder.

Step 3: Plotting the traces

Finally, you can plot the traces by running the plot.py script. If everything worked, there should be obvious differences between the plots of the Linux and OpenSSL implementations.

Note

The files src/victims/aesni-linux.S and src/victims/aesni-openssl.s contain code taken from Linux and OpenSSL, and are published under their respective licenses.