Secure Machine Learning Classification - secMLClass, the source code of a unified framework for secure Machine Learning (ML) classifications over encrypted data for ML classifiers that can be expressed in terms of matrix and linear algebra operations such as Naive Bayes (NB), Multinomial Naive Bayes (MNB), Logistic Regression (LR), Support Vector Machines (SVM) and Deep Neural Networks (DNN). secMLClass was proposed in the confenrece paper "Secure Matrix Operations for Machine Learning Classifications Over Encrypted Data in Post Quantum Industrial IoT" presented at the IEEE 2021 International Symposium on Networks, Computers and Communications (ISNCC'21) held in Dubai/UAE from October 31 to November 2, 2021. Based on where the bulk of operations are done, for all of those classifiers, secMLClass algorithm provides both the server and the client/user centric scenarious.
For benchmark/comparison purposes with the related privacy-preserving state of the art schemes, we use 3 datasets:
- Breast Cancer Wisconsin (original) dataset, which can be found at https://archive.ics.uci.edu/ml/datasets/breast+cancer+wisconsin+(original)
- The ADFA-NB15 cyber-security dataset, which can be found at https://www.unsw.adfa.edu.au/unsw-canberra-cyber/cybersecurity/ADFA-NB15-Datasets/
- The enron e-mail dataset, generated by - Metsis, Vangelis, Ion Androutsopoulos, and Georgios Paliouras. "Spam filtering with naive bayes-which naive bayes?." CEAS. Vol. 17. 2006.
Link to the paper: Coming soon
Video presentation of the paper: https://www.youtube.com/watch?v=4d3G5AmfbHU&list=PLN2gEfNq4GvfUYJx1aBE7R8bIAXt4JK2P&index=2
The datasets are already included with the uploaded secMLClass.rar package of the project. Just download & extract it and open the solution using Visual Studio 2017 or any other IDE of your choice. Set the SEALExamples as the start-up project and, for performance reasons, run it in Release mode. The implementation uses Microsoft's SEAL library v3.2 found in https://github.com/microsoft/SEAL
For any inqueries you can contact me by artrimq@gmail.com or artrimk@sabanciuniv.edu