OGHE - Oncological Genomic analysis over HE

The repository contains the code needed to replicate the experiments presented in "Enhancing privacy-preserving cancer classification with Convolutional Neural Networks", which will be published in the 2025 Pacific Symposium on Biocomputing. Dataset is available at https://drive.google.com/drive/folders/16taBYFtPdygWOZAS60Y6L3z9Z-YGSoIq?usp=sharing .

In particular:

• .devcontainer contains the files needed to create a devcontainer using VSCode;
• Dataset contains the iDash2020 dataset;
• requirements.txt is the Python requirements file. Install all the required libraries with pip install -r requirements.txt;
• data_utils.py contains the pre-processing functions, needed to parse and load the dataset;
• experiment_*_shuffle.py contains the code of the FC, NN, and OGHE with the data pre-processing of Hong et al.;
• experiment_*_shuffle_OurPreProcessing.py contains the code of the FC, NN, and GenHECNN with our proposed pre-processing;
• StatisticalTest contains the code to run the statistical tests of difference between the proposed solution and Hong et al.;
• Timing contains the code for the encrypted computation time. In particular, to run the timing experiments, run:

Encrypted experiments

To run the encrypted tests, run

python3 Test_single.py --n_jobs=1 --n_samples=1

for a single sample; note that it will also output the final content of the ciphertexts to check the correctness vs the plain processing, and

python3 Test_multithread.py --n_jobs 40 --n_samples 100

for multiple samples.