From: Confidential machine learning on untrusted platforms: a survey
Framework | ML Algorithm | Original Component | Crypto-friendly Component | Benefits |
---|---|---|---|---|
Logistic Regression, Neural Networks | Sigmoid, Softmax | ReLu | Avoids inversion and limits expensive confidential divisions to one. | |
LMC, Fisher’s LDA | Divisions | Multiplications with incorporated division factors | Avoids division costs and simplifies the protocol. | |
Ridge Linear Regression | LU decomposition | Cholesky’s decomposition | Reduces the cost complexity by half. | |
Matrix Factorization | Cholesky’s Decomposition | Sorting based matrix factorization | Reduces the overall complexity from quadratic to within a polylogarithmic factor of the complexity in the plaintext | |
Boosting | Decision Stumps | Random Linear Classifiers | Reduced number of comparisons and simplicity in learning. | |
Logistic Regression | Exponentiation | Taylor Expansion | Avoids costs involved in multiple levels of multiplications. | |
Spectral Clustering | Eigen decomposition | Eigen-approximation by Lanczos and Nystrom | Reduces complexity of the problem from O(N3) to O(N2). |