Reproduce “Secure Change-Point Detection for Time Series under Homomorphic Encryption” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Federico Mazzone, Giorgio Micali, Massimiliano Pronesti
Researchers at the University of Twente and IBM Research Europe present the first fully encrypted method for change-point detection in time series using homomorphic encryption, achieving plaintext-level accuracy on synthetic and real-world datasets. The system processes one million data points for frequency shift detection in approximately three minutes, demonstrating practical feasibility for secure outsourced time series analysis.
Key insights
- Homomorphic encryption can enable accurate change-point detection on time series without sacrificing data utility, a limitation of differential privacy approaches.
- Efficient homomorphic implementations for non-parametric statistics, such as turning rates based on ordinal patterns, allow for robust detection on dependent time series.
- Algorithmic optimizations for partial sums and argmax operations significantly reduce the computational overhead and circuit depth of complex homomorphic analytical tasks, improving practical feasibility.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.